Search

The fish industry is an important source of income for island countries. Fish is a main source of animal-based protein. Marine fishing is gradually being replaced by marine farming (or aquaculture) due to declining wild fish populations and water pollution. However, fish farming is costly job with high requirements for labor, electricity, water, and feed. The use of deep learning to perform intelligent surveillance in aquaculture fields, reducing the need for human resources and implementing real-time monitoring, has been proposed. In this article, we propose a novel deep residual network (ResNeXt $3 \times 1 \mathrm{D}$) for abnormal behavior recognition in aquaculture fields. The proposed ResNeXt $3 \times 1 D$ convolutional network is mainly based on an $R(2+1) D$ convolutional network and modified to obtain better performance. Experimental results showed that the proposed ResNeXt $3 \times 1 D$ exhibited good performance for abnormal behavior recognition in aquaculture fields. Specifically, the accuracy obtained using the proposed ResNeXt $3 \times 1 \mathrm{D}$ for abnormal behavior recognition in aquaculture fields was approximately $95.3 \%$. [ABSTRACT FROM AUTHOR]

Unmanned aerial vehicles (UAVs) foster the development of multifield applications and also present opportunities for malicious activities, and thus, comprehensive surveillance of UAVs is indispensable. Taking advantage of rapidly developing communication technology, perceptive mobile networks (PMNs) are emerging to integrate sensing capability into cellular networks, which paves the way for ubiquitous networked sensing with cooperation among interconnected base stations (BSs). This article explores key issues in designing PMNs for reliable UAV surveillance, e.g., sensing coverage, network clutter, and the performance tradeoff between sensing and communication. Procedures and key strategies of cooperative sensing in PMNs are also presented, highlighting the method of conducting cooperative sensing by multiple BSs to reap spatial diversity gain. Finally, we depict our vision for future research directions of cooperative sensing in PMNs. [ABSTRACT FROM AUTHOR]

In this letter, a cost-effective sensing-capable multihop router network architecture has been designed for the hotel indoor surveillance system with the help of wemos D1 R2, Arduino Mega, relays, and sensors (WARS). To make each router, there are mainly three components, i.e., one access point (AP), one station (STA), and one controller. These components have different working functionalities. Though we have focused on hotel surveillance systems but the network will be useful for broad application purposes, including home, hospital, restaurant, agriculture, etc., according to the requirement of sensors. To regulate the information of the sensor, a smartphone application (APP) has been designed. With this architecture, the proposed multihop network can increase the regularity performance of the hotel without violating customers’ privacy. [ABSTRACT FROM AUTHOR]

A surveillance system using object tracking enhances security by continuously monitoring and analyzing movements. However, tracking human movement in surveillance systems still presents challenges, especially in crowded environments. These challenges, such as occlusion, similar appearance, and deformation, can affect the accuracy and precision of object tracking. To address these issues, we introduced a new approach combining HSV-template matching with the MEESPSO algorithm. In this approach, HSV-template matching continuously detects the target object in sequence images, while the Midrange Exploration Exploitation Searching Particle Swarm Optimization (MEESPSO) algorithm searches for the target location in the frames. The proposed method was tested using six video datasets from Performance Evaluation of Tracking and Surveillance 2009 (PETS09) and Multiple Object Tracking Challenge 2020 (MOT20), encountering crowded environments that introduced challenges like occlusion, similar appearance, and deformation. Experiments on both PETS09 and MOT20 datasets demonstrated that the proposed method improved tracking performance by over 4.67% in accuracy and 15% in precision compared to existing studies, where the result effectively addresses the crowded environment challenges identified in this research.

Biometric recognition has become essential for secure and reliable access control in high-security systems such as surveillance, law enforcement, and smart cities. While deep learning models offer exceptional performance in biometric recognition, they are susceptible to security challenges such as adversarial attacks. Current research addresses the vulnerability of single modalities systems. However, there is a gap in understanding the impact of adversarial attacks on fusion levels in multimodal biometric systems. This research aims to fill this gap by thoroughly assessing the vulnerability of multimodal biometric systems to adversarial attacks to enhance their security in real-world high-security applications. This research investigates the most secure fusion level for combining behavioral and biological biometric traits in multimodal biometric authentication systems under adversarial attacks. We assess the security of different fusion levels by employing the Fast Gradient Sign Method (FGSM). Also, our study contributes to the field by evaluating the extent of perturbations needed to generate effective adversarial attacks and identifying the fusion level that offers the highest security under different perturbations. Our experiments thoroughly analyze multimodal fusion levels using attack success rate, accuracy, precision, and recall under clean and adversarial data conditions. According to our results, the input fusion level offers the most secure level among the three fusion levels. In various adversarial attack settings, it demonstrates an average attack success rate of 16.62% on DenseNet201 and 32.30% on ArcFace architectures. This research presents an important analysis to support further investigations into the security of multimodal biometric systems and building defense methods for such systems.

The rapid development of self-driving vehicles requires integrating a sophisticated sensing system to address the various obstacles posed by road traffic efficiently. While several datasets are available to support object detection in autonomous vehicles, it is crucial to carefully evaluate the suitability of these datasets for different weather conditions across the globe. In response to this requirement, we present a novel dataset named the Canadian Vehicle Datasets (CVD). Subsequently, we present deep learning models that use this dataset. The CVD comprises street-level videos which were recorded by Thales, Canada. These videos were collected with high-quality cameras mounted on a vehicle in the Canadian province of Quebec. The recordings were made during daytime and nighttime, capturing weather conditions such as hazy, snowy, rainy, gloomy, nighttime and sunny days. A total of 10000 images of vehicles and other road assets are extracted from the collected videos. A total of 8388 images were annotated with corresponding generated labels 27766 with their respective 11 different classes. We analyzed the performance of the YOLOv8 model trained using the existing RoboFlow dataset. Then, we compared it with the model trained on the expanded version of RoboFlow using the proposed weather-specific dataset, CVD. Final values of improved accuracy of 73.26 %, 72.84 %, and 73.47 % (Precision/Recall/mAP) were reported upon adding the proposed dataset. Finally, the model trained on this diverse dataset exhibits heightened robustness and proves highly beneficial for both autonomous and conventional vehicle operations, making it applicable not only in Canada but also in other countries with comparable weather conditions.

Convolutional neural networks have demonstrated remarkable capability in extracting deep semantic features from images, leading to significant advancements in various image processing tasks. This success has also opened up new possibilities for change detection (CD) in remote sensing applications. But unlike the conventional image recognition tasks, the performance of AI models in CD heavily relies on the method used to fuse the features from two different phases of the image. The existing deep-learning-based methods for CD typically fuse features of bitemporal images using difference or concatenation techniques. However, these approaches often fail tails to prioritize potential change areas adequately and neglect the rich contextual information essential for discerning subtle changes, potentially leading to slower convergence speed and reduced accuracy. To tackle this challenge, we propose a novel feature fusion approach called feature-difference attention-based feature fusion CD network. This method aims to enhance feature fusion by incorporating a feature-difference attention-based feature fusion module, enabling a more focused analysis of change areas. Additionally, a deep-supervised attention module is implemented to leverage the deep surveillance module for cascading refinement of change areas. Furthermore, an atrous spatial pyramid pooling fast is employed to efficiently acquire multiscale object information. The proposed method is evaluated on two publicly available datasets, namely the WHU-CD and LEVIR-CD datasets. Compared with the state-of-the-art CD methods, the proposed method outperforms in all metrics, with an intersection over union of 92.49% and 85.56%, respectively. [ABSTRACT FROM AUTHOR]

Several prominent criticisms have recently challenged the possibility of algorithmically determining or recognising human affect. This paper ethically evaluates one underexplored avenue for overcoming such deficiencies in categorical affect recognition technologies (ARTs). Specifically, the emerging literature on “situated affectivity” offers valuable guidance on three fronts. First, it conceptually substantiates existing criticism by stressing the contextual dimensions of human affect. Second, in highlighting the breadth of information required for more accurate ARTs, it suggest that they may necessitate penetrating forms of surveillance the ethical literature has already provided good reason to resist. Third, it corroborates postphenomenological insights concerning the socially co-constituting character of technology; and thereby grounds the need for a “situated ethics” sensitive to the ways that ARTs might reconstruct human affectivity (and reinforce existing power hierarchies). The deep worry is that this concern is likely to persist even if concerns over accuracy and surveillance are mitigated. A detailed consideration of this novel ethical concern suggests that the ethically acceptable use of ARTs may only be possible in highly contained settings. Yet doubts over such containment places advocates of ARTs in a “double bind” they need to address. [ABSTRACT FROM AUTHOR]

Compressive sensing (CS) theory have been proposed in the field of radar for target detection. The challenge of CS applied to passive bistatic radar (PBR) lies in the high-computational complexity aggravated by the transmitted signal dependent time-varying sparse model. So, we propose a transmitted-signal-free and time-invariant sparse model for passive radar based on orthogonal frequency division multiplexing waveforms. We first generate the sparse model using the surveillance signal and pilot information only, and exploiting the sparsity of scene including only a few targets and clutter. Then, range-Doppler profile for target detection can be implemented based on our proposed sparse model. Finally, simulation and experimental results illustrate that our proposed sparse model has high detection performance. [ABSTRACT FROM AUTHOR]

This letter presents a radar waveform design for deceiving noncooperative bistatic radars (NCBRs). The principle of deceiving the noncooperative bistatic radar is proposed according to some imperfections in the engineering realization of NCBRs whose direct wave parameters are selected from the template library to reconstruct the reference signal. These imperfections are either unavoidable, such as phase asynchrony, or hard to find, such as time errors. For the first time, a radar waveform design method based on joint modulation of the initial phase and the pulse repetition interval (PRI) is proposed to achieve range migration–velocity deception for NCBRs. When our host radar illuminates our key target, the noncooperative bistatic radar will detect our key target. The proposed waveform will mislead the target detection results of the noncooperative bistatic radar in range–time (R–T) maps and range-Doppler (R-D) maps. It can not only be used to disguise a moving target as a stationary point target to avoid moving target detection (MTD) of NCBRs, but also modulate the target echo velocity into the opposite direction of the real motion to prevent our key moving targets from being tracked. Both the simulation data and the measured results show that the deception method proposed in this letter fulfills the requirements of NCBR countermeasures [ABSTRACT FROM AUTHOR]

Researchers who have worked in computer science or artificial intelligence for more than a few years will have experienced a profound shift in the public perception of their field from being something that people rarely cared about outside of sci-fi or niche academic or “geek” interest to something that is suddenly everywhere all over the world and seems to be permeating every field and every product. How should we respond to this? [ABSTRACT FROM AUTHOR]

Unmanned Aerial Vehicles (UAVs), sometimes known as drones, evolved from military to civilian applications, opening up novel perspectives in a variety of everyday services. The rapidly growing consumer interest in amateur drones equipped with high-end cameras compromises the everyday safety and privacy of people. In the literature, a variety of sensing techniques based on different physical phenomena have been proposed for drone detection. Among acoustic, optical, or radar detection systems, passive radiofrequency sensing is the only one that can identify a drone even before it takes off and additionally indicate the operator’s location. A spectrogram-based method is developed and optimised in terms of computing location, resulting in the possibility of sensor grid deployment over a standard Ethernet network. The detection phase involves hardware-accelerated energy sensing to extract the data frames from the background noise. Drone presence is then identified using machine learning based solely on preamble pattern recognition, which reduces the computational effort. The presented procedure is evaluated in an isolated setting employing an open-source dataset and tuned across multiple neural network architectures. Next, the complete sensor processing chain is examined in a real-life scenario. The analytical energy detector stage reaches a margin of roughly −8.7 dB in the signal-to-noise (SNR) ratio. With 1.1 M parameters, the proposed neural network achieves 99.93% simulation accuracy in up to −9.5 dB SNR range. Even after quantization for embedded platform implementation, the device can be used as a stand-alone early intrusion detector or as part of a distributed sensor grid.

Blockchain (BC) and Computer Vision (CV) are the two emerging fields with the potential to transform various sectors. BC can offer decentralized and secure data storage, while CV allows machines to learn and understand visual data. The integration of the two technologies holds massive promise for developing innovative applications that can provide solutions to the challenges in various sectors such as supply chain management, healthcare, smart cities, and defense. This review explores a comprehensive analysis of the integration of BC and CV by examining their combination and potential applications. It also provides a detailed analysis of the fundamental concepts of both technologies, highlighting their strengths and limitations. This paper also explores current research efforts that make use of the benefits offered by this combination. The BC can be used as an added layer of security in CV systems and also ensure data integrity, enabling decentralized image and video analytics. The challenges and open issues associated with this integration are also identified, and appropriate potential future directions are also proposed.

Vehicle re-identification is an important feature of an intelligent transportation system as part of a smart city application. Vehicle re-identification aims at matching vehicles from images acquired by surveillance cameras at different locations. During rush hours, vehicles are densely occupied across regions such as entry/exit of gated campuses, railways, airports, educational institutions, etc. Due to this uneven flow of traffic, there is a possibility of violation of traffic rules by the vehicles that lead to a security breach. In such scenarios to speed up the re-identification process, it is justified to look into a specific group of surveillance cameras to detect and re- identify vehicles on day to day basis in near real-time. However, the existing vehicle re-identification datasets do not contain zone specific information and therefore can not be used to evaluate the performance of re-identification algorithms in different zones. In the proposed work for re-identification, a framework is developed that performs vehicle re-identification across a group of cameras that monitors varying traffic movements over an area. These areas defined as “strategic zones” comprise a subset of non-overlapping cameras that are installed to monitor non-uniform vehicle movements. The re-identification framework is evaluated on a novel dataset developed to understand the performance of vehicle re-identification across strategic zones. The dataset consists of videos of vehicles captured through 20 CCTV surveillance cameras that are grouped into four different zones. Various experiments are conducted to study the performance of re-identification across four zones using a deep neural network with triplet loss, L2 regularization, and re- ranking. The experiments conducted with an image dimension of $224\times224$ have demonstrated an overall mAP of 77.22%. Also, for each of the four zones a mAP of 82.16%, 69.1%, 66.5%, and 75.76% is achieved. The experimental results demonstrate huge variations in the accuracy of vehicle re-identification method across different zones. Therefore, the study assess the possible measures that can be taken to improve the performance in individual zones for an accurate vehicle re-identification in intelligent transport system.

In modern networking research, infrastructure-assisted unmanned autonomous vehicles (UAVs) are actively considered for real-time learning-based surveillance and aerial data-delivery under unexpected 3D free mobility and coordination. In this system model, it is essential to consider the power limitation in UAVs and autonomous object recognition (for abnormal behavior detection) deep learning performance in infrastructure/towers. To overcome the power limitation of UAVs, this paper proposes a novel aerial scheduling algorithm between multi-UAVs and multi-towers where the towers conduct wireless power transfer toward UAVs. In addition, to take care of the high-performance learning model training in towers, we also propose a data delivery scheme which makes UAVs deliver the training data to the towers fairly to prevent problems due to data imbalance (e.g., huge computation overhead caused by larger data delivery or overfitting from less data delivery). Therefore, this paper proposes a novel workload-aware scheduling algorithm between multi-towers and multi-UAVs for joint power-charging from towers to their associated UAVs and training data delivery from UAVs to their associated towers. To compute the workload-aware optimal scheduling decisions in each unit time, our solution approach for the given scheduling problem is designed based on Markov decision process (MDP) to deal with (i) time-varying low-complexity computation and (ii) pseudo-polynomial optimality. As shown in performance evaluation results, our proposed algorithm ensures (i) sufficient times for resource exchanges between towers and UAVs, (ii) the most even and uniform data collection during the processes compared to the other algorithms, and (iii) the performance of all towers convergence to optimal levels.

The detection of moving objects in airport apron surveillance encounters serious detection defects, which reflects that the classification is biased toward the background. In this article, we propose a new method, scene-specific-prior–based moving object detection (SMod), to detect the moving object by the use of the prior knowledge specific to the airport apron. We find that the airport apron is made of concrete and has a special gray–white color distribution. Therefore, first, we use a dual-mode Gaussian distribution to fit the color distribution of the airport apron. Further analysis shows that this fitted distribution also reflects the probability of a detection defect, which occurs when the aircraft and occluded apron area share a similar color distribution. Based on this, second, we present a scene-specific prior model based on the dual-mode distribution, where pixels with a larger detection defect probability are more likely to be classified as the foreground, so that the prior model is able to bias the classification to the foreground. As a result, the detection defect in the airport apron can be compensated for. Third, the prior model is used in conjunction with an apron model and an object model to detect a moving object within a Bayesian framework. Next, we conduct comparison experiments on the airport ground video surveillance benchmark to verify the effectiveness of the proposed algorithm. Finally, we demonstrate the application of SMod in an airport apron surveillance system. [ABSTRACT FROM AUTHOR]

Visual surveillance for autonomous power facility inspection is considered to bethe prominent field of study in the power industry. This research field completely focuses on either object detection or image fusion which lacks the overall consideration. By considering this, a single end-to-end object detection method by incorporating the image fusion named Fast Detection Fusion Network (FDFNet) is proposed in this paper to output qualitative fused images with detection results. The parameters in the FDFnet are greatly reduced by sharing the feature extraction network between image fusion and object detection tasks, due to which a huge reduction in computational complexity is achieved. On this basis, the object detection algorithm performance on various types of power facility images is compared and analyzed. For experimentation purposes, an IR (infrared) and VIS (visible) image acquisition system has also been designed. In addition, a dataset named CVPower with different sets of images for power facility fusion detection is constructed for this research field. Experimental results demonstrate that the proposed method can achieve the mAP of not less than 70%, process 2 frames per second, and produce high qualitative fused images. [ABSTRACT FROM AUTHOR]

Entity resolution (ER) identifies data instances referring to the same real-world entity and has received enormous research attention. In this paper, we examine the task of ER from a broader perspective, with its input extended from textual records, which are conventionally studied in the literature, to other modalities such as check-in sequences, GPS trajectories and surveillance video frames to generate new applications. Our goal in this paper is to design an effective model to uniformly support all these ER applications with different input formats. Technically, we fully exploit the semantic contexts of embedding vectors for the pair of input instances. In particular, we propose an integrated multi-context attention framework that takes into account self-attention, pair-attention and global-attention from three types of context. The idea can be further extended to incorporate attribute attention in order to support structured datasets. We conduct extensive experiments on a diverse class of entity resolutions tasks, including tasks on unstructured, structured and dirty textual records, check-in sequences, GPS trajectories and surveillance video frames. The experimental results verified the effectiveness and generality of our model. When compared with strong baselines in these applications, our model can achieve superior or comparative performance. [ABSTRACT FROM AUTHOR]

Advanced air mobility (AAM) is an emerging industry focus as well as a research and development discipline. Innovations and technologies resulting from AAM will change the way that we move cargo and people in and around cities. Industry is moving fast with excitement to deploy AAM solutions. However, there are multiple technical challenges that need to be overcome before AAM becomes a reality. This article takes a closer look at the technology readiness level of AAM solutions in the area of communications, navigation, and surveillance (CNS) and identifies open research problems as well as directions to address them. In particular, we discuss current approaches and future research challenges in air corridor design, air-to-air (AA) communications, 3rd Generation Partnership Project (3GPP) support for navigation, and detect and avoid (DAA)/collision avoidance, among other areas, for supporting future AAM operations. [ABSTRACT FROM AUTHOR]

Vehicles detection via surveillance sensors is an intractable and essential task under nighttime urban intersection scenes. To efficiently resolve the problem that most of the current color or texture feature-based vehicles detection models easily suffer contamination from dynamic and sudden or gradual illumination variations on nighttime, an adaptive vector-based background model without complex artificial texture feature is proposed for vehicles detection in night intersection surveillance scenes. For each pixel of image, a series of filters are employed to take the values from the past at the same location and its neighborhood, and a vector is assigned to each filter to represent the neighborhood region information of the pixel. Then, the series of vectors comprising the background model and the vector of the current pixel is compared to the vectors of the background model based on the theorem of linear dependence to describe whether that pixel belongs to foreground or background. Eventually, to adapt the dynamic scenes, a random update scheme is employed to update the model. As our experimental results demonstrated in real-world nighttime urban intersection surveillance scenarios, the proposed model attains superior vehicles detection performance compared to other state-of-the-art algorithms based on extensive qualitative and quantitative evaluations. [ABSTRACT FROM AUTHOR]

Drones, or unmanned aerial vehicles (UAVs), are among the most beneficial and emerging technologies, with a wide range of applications that can support the sustainability concerns of smart cities and ultimately improve citizens’ quality of life. The goals of this systematic review were to explore the involvement of surveillance drones in smart cities in terms of application status, application areas, proposed models, and characteristics of drones. We conducted this systematic review based on the preferred reporting items for systematic reviews and meta-analyzes (PRISMA) guidelines. We systematically searched the Web of Science and Scopus for journal articles and conference papers written in English and published up to August 2021. Of the 323 records identified, 43 met the inclusion criteria. Findings showed that surveillance drones were used in seven distinct research fields (transportation, environment, infrastructure, object or people detection, disaster management, data collection, and other applications). Air pollution and traffic monitoring were the dominant application areas. The majority of reviewed models were based on the application of rotary-wing single-drones with the camera as the aerial sensor. Reviewed models showed that the adoption of a single or multiple UAVs, either as a stand-alone technology or integrated with other technologies (e.g., internet of things, wireless sensor networks, convolutional neural networks, artificial intelligence, machine learning, computer vision, cloud computing, web applications), can offer efficient and sustainable solutions compared to conventional surveillance methods. This review can benefit academic researchers and practitioners.

Radio Frequency (RF) spectrum monitoring and broadband signal analysis have multiple application areas, especially in the era of a constantly growing number of wireless devices. One of the essential challenges for a spectrum sensor is to achieve an adequate measurement rate over a wide bandwidth to detect signals of short duration so that a low latency response can be provided. In procedures that require field measurements, and some compromise in accuracy is acceptable, low-cost Software Defined Radio (SDR) devices can be used instead of expensive and bulky professional spectrum analyzers. This paper introduces a real-time swept spectrum sensor based on LimeSDR-USB with custom embedded Field Programmable Gate Array (FPGA) firmware, designed to outperform similar software implementations. The Welch’s spectral density estimation is implemented in hardware to minimise the USB transfer rate and offload the host PC signal processing. Furthermore, the frequency tuning state machine and cache calibration memory are also managed by the FPGA to reduce the blind time during broadband sweep. The performance of the proposed solution indicates up to 96 MHz of real-time bandwidth along with a capability of less than millisecond cumulative sweep time per gigahertz. The characteristics of various design elements are investigated and refined during simulation and laboratory measurements, whereas the final prototype implementation is verified in real-world scenarios. The results demonstrate the effectiveness of the proposed device as a sensor for propagation studies, multiband spectrum utilisation monitoring, and spectral white spaces detection.

Human action recognition (HAR) technology is receiving considerable attention in the field of human-computer interaction. We present a HAR system that works stably in real-world applications. In real-world applications, the HAR system needs to identify detailed actions for specific purposes, and the action data includes many variations. Accordingly, we conducted three experiments. First, we tested our recognition system’s performance on the UTD-MHAD dataset. We compared our system’s accuracy with results from previous research and confirmed that our system achieves a 91% average performance among recognition systems. Furthermore, we hypothesized the use of a HAR system to detect burglary. In the second experiment, we compared the existing benchmark data with our crime detection dataset. We recognized the test scenarios’ data by using the recognition system trained by each dataset. The recognition system trained by our dataset achieved higher accuracy than the past benchmark dataset. The results show that the training data should contain detailed actions for a real application. In the third experiment, we tried to find the motion data type that stably recognizes action regardless of data variation. In a real application, the action data are changed by people. Thus, we introduced variations in the action data using the cross-subject protocol and moving area setting. We trained the recognition system using each position and angle data. In addition, we compared the accuracy of each system. We found that the angle format results in better accuracy because the angle data are beneficial for converting the action variation into a consistent pattern.

Drones, or general UAVs, equipped with cameras have been fast deployed with a wide range of applications, including agriculture, aerial photography, and surveillance. Consequently, automatic understanding of visual data collected from drones becomes highly demanding, bringing computer vision and drones more and more closely. To promote and track the developments of object detection and tracking algorithms, we have organized three challenge workshops in conjunction with ECCV 2018, ICCV 2019 and ECCV 2020, attracting more than 100 teams around the world. We provide a large-scale drone captured dataset, VisDrone, which includes four tracks, i.e., (1) image object detection, (2) video object detection, (3) single object tracking, and (4) multi-object tracking. In this paper, we first present a thorough review of object detection and tracking datasets and benchmarks, and discuss the challenges of collecting large-scale drone-based object detection and tracking datasets with fully manual annotations. After that, we describe our VisDrone dataset, which is captured over various urban/suburban areas of 14 different cities across China from North to South. Being the largest such dataset ever published, VisDrone enables extensive evaluation and investigation of visual analysis algorithms for the drone platform. We provide a detailed analysis of the current state of the field of large-scale object detection and tracking on drones, and conclude the challenge as well as propose future directions. We expect the benchmark largely boost the research and development in video analysis on drone platforms. All the datasets and experimental results can be downloaded from https://github.com/VisDrone/VisDrone-Dataset. [ABSTRACT FROM AUTHOR]

We introduce the problem of multi-camera trajectory forecasting (MCTF), which involves predicting the trajectory of a moving object across a network of cameras. While multi-camera setups are widespread for applications such as surveillance and traffic monitoring, existing trajectory forecasting methods typically focus on single-camera trajectory forecasting (SCTF), limiting their use for such applications. Furthermore, using a single camera limits the field-of-view available, making long-term trajectory forecasting impossible. We address these shortcomings of SCTF by developing an MCTF framework that simultaneously uses all estimated relative object locations from several viewpoints and predicts the object's future location in all possible viewpoints. Our framework follows a Which-When-Where approach that predicts in which camera(s) the objects appear and when and where within the camera views they appear. To this end, we propose the concept of trajectory tensors: a new technique to encode trajectories across multiple camera views and the associated uncertainties. We develop several encoder-decoder MCTF models for trajectory tensors and present extensive experiments on our own database (comprising 600 hours of video data from 15 camera views) created particularly for the MCTF task. Results show that our trajectory tensor models outperform coordinate trajectory-based MCTF models and existing SCTF methods adapted for MCTF. Code is available from: https://github.com/olly-styles/Trajectory-Tensors. [ABSTRACT FROM AUTHOR]

Modern methods for counting people in crowded scenes rely on deep networks to estimate people densities in individual images. As such, only very few take advantage of temporal consistency in video sequences, and those that do only impose weak smoothness constraints across consecutive frames. In this paper, we advocate estimating people flows across image locations between consecutive images and inferring the people densities from these flows instead of directly regressing them. This enables us to impose much stronger constraints encoding the conservation of the number of people. As a result, it significantly boosts performance without requiring a more complex architecture. Furthermore, it allows us to exploit the correlation between people flow and optical flow to further improve the results. We also show that leveraging people conservation constraints in both a spatial and temporal manner makes it possible to train a deep crowd counting model in an active learning setting with much fewer annotations. This significantly reduces the annotation cost while still leading to similar performance to the full supervision case. [ABSTRACT FROM AUTHOR]

This article introduces a new persistent surveillance problem, which is to detect events that randomly occurred on a road network by cooperative unmanned ground vehicles (UGVs) with a certain detection ability, where the road network is simplified into weighted viewpoints and path segments. To tackle the road network cooperative persistent surveillance problem, this article seeks to: 1) evaluate the surveillance effect, and 2) plan the trajectories of UGVs to achieve the optimal surveillance effect. First, the nonconfidence level of event occurrence status at the viewpoint, or the uncertainty, which is derived from the detection ability, the monitoring interval, and the weight of viewpoints, is defined as a performance criterion to evaluate the surveillance effect. Then, the new persistent surveillance problem is formulated as a combinatorial optimization problem and proved to be NP-hard. Considering the characteristics of the problem, a heuristic cooperation path planning algorithm with polynomial time complexity is designed to generate noncyclic trajectories. Theoretical analysis and simulation results demonstrate the feasibility of the proposed algorithm on road networks of any topology. By comparing with the existing methods in simulations and outdoor experiments, it demonstrates that the proposed algorithm has the superiority of reducing the uncertainty of the road network and improving the surveillance efficiency in real-world applications. [ABSTRACT FROM AUTHOR]

The rapid development of urbanization has raised challenges for existing parking facilities to serve increasing parking demand. Being constrained by the limited urban land resources for newly constructed parking facilities, improving the efficiency of existing parking infrastructures relies more on an advanced parking management strategy. Video-based parking surveillance technology, with abundant information, easy installation, and powerful algorithms, is the most popular method to provide real-time parking information as the required data feed for a parking management system. Meanwhile, deep learning-based algorithms gradually replace traditional computer vision methods for video processing since the detection accuracy has been increased considerably by learning fine-grained features. However, the excessive computational complexity of deep learning-based algorithms occupies considerable computational resources, which certainly hurts the entire system’s efficiency. Due to the limited computing power of edge devices, most parking surveillance systems deploy video processing algorithms on a server or cloud platform, which raises concerns about data transmission latency and central computation pressure. Deploying efficient algorithms on an edge-side device is a potential solution to solve these problems. This article proposes an edge computing parking occupancy detection system with a quantized deep learning model. Model quantization is employed to boost the inference speed while maintaining accuracy. In addition, knowledge distillation is applied to improve the quantized model’s training performance. Experiments are conducted to demonstrate the model’s superiority compared to state-of-the-art algorithms and the feasibility of edge computing. The proposed method can improve the accuracy and efficiency of parking surveillance systems. It is a systematic solution for obtaining parking information with limited computational resources. [ABSTRACT FROM AUTHOR]

Asan important means in low-altitude airspace surveillance, passive radar has been widely studied. With the increasing complexity of the low-altitude environment, the urgent need for radar's capability of automatic target recognition (ATR) emerges. In the field of ATR for passive radar, the method based on radar cross section (RCS) is an important direction for the simplicity of feature extraction. In this article, we exploit the compensated quasi-echo-power (CQEP), i.e., quasi-RCS, to realize ATR in passive radar. We propose an ATR method on the basis of the angular diversity. First, we divide the angular space, which consists of four components of the incident angle and scattering angle of electromagnetic wave, into multiple subspaces. Then, we build a subrecognizer in each angular subspace according to the CQEP distribution characteristics. Each receiving station of the multistatic passive radar recognizes targets in an angular subspace corresponding to its observation and outputs a preliminary decision. Based on the preliminary decisions of all receiving stations, a voting strategy is adopted to make the final decision on the unknown target class in the fusion center. Experimental results based on real-life target data show that the proposed method has a significant improvement on the average correct recognition rate with fusion of three receiving stations, which validates the proposed method. [ABSTRACT FROM AUTHOR]

This article considers the detection and tracking of weak targets in multiple pulse repetition frequency (multi-PRF) radars using the track-before-detect (TBD) technique. By exploring the measurement independence among different PRFs, we decompose the joint multi-PRF and multiframe optimization problem into several lower dimensional maximizations, each of which corresponds to an intra-PRF multiframe processing. An efficient two-stage TBD procedure with a parallel structure is proposed for the multi-PRF radars. Specifically, in the first stage, a constraint inequality is derived analytically and used to decouple the measurement ambiguities from the nonlinear conversion relationship between polar and Cartesian coordinates. The intra-PRF multiframe integration can then be carried out concurrently using the ambiguous measurements and the target-like measurement plot sequences with different ambiguities are extracted for different PRFs. In the second stage, a covariance combination fusion-based inter-PRF joint disambiguation and estimation algorithm is proposed to solve the ambiguities of the plot sequences and output high-accuracy target tracks. Simulation experiments show that the proposed algorithm can provide a good detection performance and higher tracking accuracy with much lower complexity. [ABSTRACT FROM AUTHOR]

Recently, human motion-centric videos have been attracting attention in the field of computer vision. Observing and detecting human motion in intelligent surveillance camera systems is essential for understanding the intentions of target subjects. However, these videos have vast amounts of disparate and complex information, and hence they are difficult to process and label automatically. As a result, building and maintaining a database using motion-centric videos requires considerable labor in trimming and classifying the videos. Therefore, we propose a self-updatable motion database system based on a human motion assessment framework for evaluating complex human movements. The framework quantifies three primitive motion properties: stability, liveliness, and attention. This assessment highlights the semantics of human motion in the input video. The semantic motion sequence obtained after the motion assessment is compared with a similarity motion database to determine whether the database needs to be updated; for efficient comparison, we introduce a sequential autoencoder model with a long short-term memory neural network. The proposed system maintains the database within a surveillance camera system using a motion update algorithm; unseen motions in the database are updated using a camera-based surveillance system. In addition, this framework combines state-of-art action recognition methods to improve performance by up to 11% via the self-update of motion. [ABSTRACT FROM AUTHOR]

Drone-based vehicle detection aims at detecting vehicle locations and categories in aerial images. It empowers smart city traffic management and disaster relief. Researchers have made a great deal of effort in this area and achieved considerable progress. However, because of the paucity of data under extreme conditions, drone-based vehicle detection remains a challenge when objects are difficult to distinguish, particularly in low-light conditions. To fill this gap, we constructed a large-scale drone-based RGB-infrared vehicle detection dataset called DroneVehicle, which contains 28, 439 RGB-infrared image pairs covering urban roads, residential areas, parking lots, and other scenarios from day to night. Cross-modal images provide complementary information for vehicle detection, but also introduce redundant information. To handle this dilemma, we further propose an uncertainty-aware cross-modality vehicle detection (UA-CMDet) framework to improve detection performance in complex environments. Specifically, we design an uncertainty-aware module using cross-modal intersection over union and illumination estimation to quantify the uncertainty of each object. Our method takes uncertainty as a weight to boost model learning more effectively while reducing bias caused by high-uncertainty objects. For more robust cross-modal integration, we further perform illumination-aware non-maximum suppression during inference. Extensive experiments on our DroneVehicle and two challenging RGB-infrared object detection datasets demonstrated the advanced flexibility and superior performance of UA-CMDet over competing methods. Our code and DroneVehicle will be available: https://github.com/VisDrone/DroneVehicle. [ABSTRACT FROM AUTHOR]

Unmanned aerial vehicles (UAVs) are increasingly applied to surveillance tasks, thanks to their excellent mobility and flexibility. Different from existing works using UAVs for video surveillance, this paper employs a UAV team to carry out collaborative radio surveillance on ground moving nodes and disguise the purpose of surveillance. We consider two aspects of disguise. The first is that the UAVs do not communicate with each other (or the ground nodes can notice), and each UAV plans its trajectory in a decentralized way. The other aspect of disguise is that the UAVs avoid being noticed by the nodes for which a metric quantifying the disguising performance is adopted. We present a new decentralized method for the online trajectory planning of the UAVs, which maximizes the disguising metric while maintaining uninterrupted surveillance and avoiding UAV collisions. Based on the model predictive control (MPC) technique, our method allows each UAV to separately estimate the locations of the UAVs and the ground nodes, and decide its trajectory accordingly. The impact of potential estimation errors is mitigated by incorporating the error bounds into the online trajectory planning, hence achieving a robust control of the trajectories. Computer-based simulation results demonstrate that the developed strategy ensures the surveillance requirement without losing disguising performance, and outperforms existing alternatives. Note to Practitioners—The paper is motivated by the covertness requirement in the radio surveillance (also called eavesdropping) by UAVs. In some situations, the UAV user (such as the police department) wishes to disguise the surveillance intention from the targets, and the trajectories of UAVs play a significant role in the disguising. However, the typical UAV trajectories such as standoff tracking and orbiting can easily be noticed by the targets. Considering this gap, we focus on how to plan the UAVs’ trajectories so that they are less noticeable while conducting effective eavesdropping. We formulate a path planning problem aiming at maximizing a disguising metric, which measures the magnitude of the relative position change between a UAV and a target. A decentralized method is proposed for the online trajectory planning of the UAVs based on MPC, and its robust version is also presented to account for the uncertainty in the estimation and prediction of the nodes’ states. [ABSTRACT FROM AUTHOR]

We present a decentralized connectivity-maintenance control framework for a heterogeneous human–robot team. The algorithm is able to manage a team composed of an arbitrary number of mobile robots (drones and ground robots in our case) and humans, for collaboratively achieving exploration and patrolling tasks. Differently from other works on the subject, here the human user physically becomes part of the team, moving in the same environment of the robots and receiving information about the team connectivity through wearable haptics or audio feedback. Although human explores the environment, robots move so as to keep the team connected via a connectivity-maintenance algorithm; at the same time, each robot can also be assigned with a specific target to visit. We carried out three human subject experiments, both in virtual and real environments. Results show that the proposed approach is effective in a wide range of scenarios. Moreover, providing either haptic or audio feedback for conveying information about the team connectivity significantly improves the performance of the considered tasks, although users significantly preferred receiving haptic stimuli w.r.t. the audio ones. Note to Practitioners—Exploration, patrolling, and search-and-rescue are highly dynamic and unstructured scenarios. When considering the operative conditions of such environments, the benefits of multirobot systems are evident. Most tasks can be carried out faster and more robustly by a team of robots with respect to a single unit. There are also situations explicitly requiring the presence of a multirobot team, e.g., using one drone for surveillance of the ground team and one ground mobile robot for carrying supplies. Of course, if the operator(s) in charge of the operation could share the same environment of the robots (i.e., be together with the robots in the field), they would be provided with a level of situational awareness that no teleoperation technology can match as of today. This work presents a framework for controlling heterogeneous teams composed of one human operator and an arbitrary number of aerial and ground mobile robots. The operator moves together with the robotic team and, at the same time, he or she receives meaningful information about the status of the formation. The algorithm only uses the relative position of the drones and humans with respect to each other, and all computations are designed in a decentralized fashion. Decentralization avoids relying on any absolute positioning system (e.g., GPS) or centralized command centers. These features make the proposed framework ready for deployment in different high-impact applications, such as in surveillance, search-and-rescue, and disaster response scenarios. [ABSTRACT FROM AUTHOR]

Vehicle Re-Identification (ReID) is of great significance for public security and intelligent transportation. Large and comprehensive datasets are crucial for the development of vehicle ReID in model training and evaluation. However, existing datasets in this field have limitations in many aspects, including the constrained capture conditions, limited variation of vehicle appearances, and small scale of training and test set, etc. Hence, a new, large, and challenging benchmark for vehicle ReID is urgently needed. In this paper, we propose a large vehicle ReID dataset, called VERI-Wild 2.0, containing 825,042 images. It is captured using a city-scale surveillance camera system, consisting of 274 cameras covering a very large area over 200 $km^2$ k m 2 . Specifically, the samples in our dataset present very rich appearance diversities thanks to the long time span collecting settings, unconstrained capturing viewpoints, various illumination conditions, diversified background environments, and different weather conditions. Furthermore, to facilitate more practical benchmarking, we define a challenging and large test set containing about 400K vehicle images that do not have any camera overlap with the training set. VERI-Wild 2.0 is expected to be able to facilitate the design, adaptation, development, and evaluation of different types of learning models for vehicle ReID. Besides, we also design a new method for vehicle ReID. We observe that orientation is a crucial factor for feature matching in vehicle ReID. To match vehicle pairs captured from similar orientations, the learned features are expected to capture specific detailed differential information for discriminating the visually similar yet different vehicles. In contrast, features are desired to capture the orientation invariant common information when matching samples captured from different orientations. Thus a novel disentangled feature learning network (DFNet) is proposed. It explicitly considers the orientation information for vehicle ReID, and concurrently learns the orientation specific and orientation common features that thus can be adaptively exploited via an adaptive matching scheme when dealing with matching pairs from similar or different orientations. The comprehensive experimental results show the effectiveness of our proposed method. [ABSTRACT FROM AUTHOR]

With the advent of rapid globalization and the inter-border supply chain network, the reliability and efficiency of transportation systems have become even more critical. Indeed, trans-continental highways need particular attention due to their important role in sustaining globalization. In this context, intelligent transportation systems (ITS) can actively enhance the safety, mobility, productivity, and comfort of trans-continental highways. However, ITS efficiency depends greatly on the roads where they are deployed, on the availability of power and connectivity, and on the integration of future connected and autonomous vehicles. To this end, high altitude platform station (HAPS) systems, due to their mobility, sustainability, payload capacity, and communication/caching/computing capabilities, are seen as a key enabler of future ITS services for trans-continental highways; this paradigm is referred to as HAPS-ITS. The latter is envisioned as an active component of ITS systems to support a plethora of transportation applications, such as traffic monitoring, accident reporting, and platooning. This article discusses how HAPS systems can enable advanced ITS services for trans-continental highways, presenting the main requirements of HAPS-ITS and a detailed case study of the Trans-Sahara highway. [ABSTRACT FROM AUTHOR]

Unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) can be jointly deployed to form a collaborative surveillance system, where UAVs collect low-resolution images at a high altitude to obtain a global perception and UGVs observe high-resolution images within a focused detection range. Such multiresolution heterogeneous observations create opportunities yet pose challenges to model the dynamics of the targeted crowds. Existing approaches that integrate multiresolution observations rely on intensive computation of a large volume of historical data, and thus, result in a lack of computational efficiency, accuracy, and robustness. To address these limitations, this article proposes a new crowd dynamics modeling approach to aggregating the multiresolution information under a Bayesian inference framework. Beta-binomial and Normal–Wishart conjugate distributions were adopted to model crowd dynamics from low-resolution UAV observations and high-resolution UGV observations, respectively. Based on the proposed approach, a real-time model updating mechanism is developed to implement information aggregation for onboard crowd dynamics inference, where high efficiency, accuracy, and robustness are critical. Numerical simulations and onboard experiments were conducted to demonstrate the effectiveness and efficiency of the proposed modeling approach. [ABSTRACT FROM AUTHOR]

This article presents an 8.3-Mpixel 60-frames/s CMOS image sensor with skew-relaxation analog-to-digital converters (ADCs) and an on- chip testable ramp generator for low-noise and high-sensitive surveillance cameras. Three novel techniques are proposed to improve the circuit noise and linearity characteristics under low illuminance. A local multiply circuit (LMC) reduces a differential non-linearity (DNL) of the column-wise ADC by halving the delay difference of long-distance wiring signals. A noise suppression ramp generator (NSRAMP) can reduce the horizontal streak noise to 1/4 by compressing only the noise component in the low-light signal. A flexible test circuit (FTC) enables high-speed on- chip test of the ramp generator separately from the ADC characteristics. Test chip fabricated in 55-nm backside illumination (BSI) process achieves DNL of +0.42 LSB/−0.32 LSB and the horizontal streak noise of 30.2 $\mu $ Vrms. DNL is reduced by 27% (circuit skew) and 50% (propagation skew), respectively, and the horizontal streak noise is also reduced by 72% compared to the conventional scheme. [ABSTRACT FROM AUTHOR]

Multiobject detection has become an integral component in various neural applications, such as autonomous driving and augmented reality. The system should be able to recognize and process multiple objects simultaneously. Moreover, the performance requirements for this system can be dynamically changed depending on the number of regions of interest (ROIs) in each frame. Consequently, the processing unit (PU) of the neural acceleration system should provide various inference rates. Therefore, we present a field-programmable gate array (FPGA)-based dynamic rate neural acceleration system called MultiLockOn to dynamically change the inference performance according to the number of ROIs per frame. It supports multiprocessing modes with different speeds through the introduction of novel multi-mode processing engines (PEs) comprising minimum reconfigurable interconnections across inference modes to minimize hardware overhead. The MultiLockOn system can provide an improvement of up to $4\times $ in the inference performance compared to that of DNNWeaver and $5.7\times $ compared to that of the ARM Cortex-A53 with minimum accuracy loss by supporting the multiprocessing modes. [ABSTRACT FROM AUTHOR]

The high-speed direct-drive blower is a high-efficiency, energy-saving, and environmentally friendly blower, and it is widely applied in industrial fields. The temperature changes significantly affect the performance of the high-speed direct-drive blower, and it is crucial to monitor its temperature. Infrared and visible light cameras are simultaneously used to measure and characterize temperature information. The measurement method can have both the temperature information of the infrared image and the contrast information of the structure and contour of the visible light image. This article establishes an image fusion representation and probabilistic generation model of infrared and visible images. Then, the information of the infrared and visible images is fused under the Bayesian framework. A hierarchical prior model using the Haar wavelet transform is proposed. The joint maximum a posteriori criterion is chosen, and an appropriate alternate optimization algorithm is designed to achieve information fusion. The proposed method is validated in industrial scenarios of high-speed direct-drive blowers. The experimental results demonstrate the robustness and effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Machine vision based vehicle re-identification (ReID) plays an important role in some Intelligent Transportation Systems (ITS). Yet, most previous methods mainly focus on fixed surveillance cameras instead of Unmanned Aerial Vehicle (UAV). With high flexibility, the UAV-based vehicle ReID problem has some special challenges including complicated shooting angles, low discrimination of top-down features, and large variance in vehicle scales, etc. To overcome these challenges, we propose a novel structure for UAV-based vehicle ReID without license plates. Firstly, a triple-head segmentation net is proposed for segmenting UAV-captured vehicles under different heights and directions. Secondly, a posture calibration model is designed to uniform the vehicle postures based on the segmentation results, with the purpose of reducing the influence of different postures. Thirdly, the novel Cross-View & Hard-Sensitive Metric Learning (CHSML) method is proposed to train a ReID network with cross-view training constraint and hard sensitive principles; the mechanism of CHSML takes the cross-view samples of same ID as a training unit to learn the potential visual relationship in cross-view and builds a hard sensitive weight matrix to make learning more focus on hard samples, which improves the low ReID accuracy brought by cross-view or hard samples. Moreover, to facilitate the research of UAV-based vehicle ReID, a large-scale UAV-based vehicle ReID dataset called VeRi-UAV is released with 17 516 vehicles of 453 IDs. The experiments show that the proposed structure gains a better performance compared with the representative methods in the UAV-based vehicle ReID task. [ABSTRACT FROM AUTHOR]

Video object segmentation (VOS) is the fundamental problem of vision-based intelligent transportation, and many VOS algorithms relying on inference from reference masks have been proposed. Due to the inherent defects of the inference strategy and the complex changes of targets, VOS methods that perform well on public datasets are usually ineffective in airport scenarios. We propose a spatiotemporal alignment network (STA-Net) that makes use of Automatic Dependent Surveillance-Broadcast (ADS-B) data as prior information to guide the long-term segmentation of aircraft. ADS-B is an airport-specific signal, which indicates the location of aircraft in real time. Based on ADS-B, we continuously generate new reference masks instead of using previous masks for inference, which greatly reduces the accumulation of inference errors. To achieve this, previous masks of each aircraft are aligned on the temporal domain based on the position information in ADS-B. All temporally-aligned masks are compared, and the one most similar to the current instant is reserved. This mask is both temporally and spatially aligned; hence it is a better reference mask for inference. Aligned masks are updated every time new ADS-B data arrive, so that they can support long-term inference. With the selected mask as a reference, aircraft of interest are segmented within a unified encoder-decoder framework over the long term. Experiments on a benchmark dataset and in a real airport scenario verify the effectiveness of the presented method. [ABSTRACT FROM AUTHOR]

CCTV-based surveillance using unmanned aerial vehicles (UAVs) is considered a key technology for security in smart city environments.This article creates a case where the UAVs with CCTV-cameras fly over the city area for flexible and reliable surveillance services. UAVs should be deployed to cover a large area while minimizing overlapping and shadow areas for a reliable surveillance system. However, the operation of UAVs is subject to high uncertainty, necessitating autonomous recovery systems. This article develops a multiagent deep reinforcement learning-based management scheme for reliable industry surveillance in smart city applications. The core idea this article employs is autonomously replenishing the UAV’s deficient network requirements with communications. Via intensive simulations, our proposed algorithm outperforms the state-of-the-art algorithms in terms of surveillance coverage, user support capability, and computational costs. [ABSTRACT FROM AUTHOR]

Traffic patterns of urban road intersections are important in traffic monitoring and accident prediction, thus play crucial roles in urban traffic management. Although real-time traffic information is consistently provided by surveillance cameras equipped at road intersections, the sparsity of surveillance distribution poses great challenges in performing a complete real-time traffic pattern analysis. To tackle that, existing works either assume that the traffic patterns are static, or assume a multi-variant distribution model for intersection traffic volumes. The former assumption neglects the temporal features of traffic patterns, and the latter is limited in capturing fine-grained spatiotemporal dependencies. To tackle the problem, we propose a novel framework, SpatioTemporal-Generative Adversarial Network (ST-GAN), that exploits deep spatiotemporal features of urban networks and offers accurate traffic pattern inferences with incomplete surveillance information. The ST-GAN framework incorporates a modified GCN network wired with the encoder-decoder mechanism and an LSTM network, which are further boosted by an iterative adversarial training process. Comprehensive experiments on real datasets show that ST-GAN achieves better inference accuracies than state-of-the-art solutions. [ABSTRACT FROM AUTHOR]

Recognizing slanted deck scenes is crucial to security monitoring for protecting ships and making them behave smartly resilient. However, there are multitude of diverse structures that are designed as slanted planes due to rough maritime environments. Traditional methods for scene understanding from 3D point clouds or RGB-D data are energy-consuming and memory intensive, which makes those models less reliable in a resource-constrained system of limited compute, memory, and energy resources on ships. In this study, we present an approach to understanding deck scenes, including slanted structures, using a low-cost monocular camera without prior training. New clusters of slanted angle projections are extracted. The vanishing points of slanted non-Manhattan angle projections are estimated. These slanted planes can be reshaped by compositions of non-Manhattan angle projections. Combined with Manhattan planes, a deck scene can be approximated by Manhattan and non-Manhattan planes. Unlike deep learning-based algorithms, this approach requires no prior training or knowledge of the camera's internal parameters. Experimental results demonstrated that the method can successfully elucidate diverse elements, including slanted structures, meeting safety monitoring requirements using a resource-constrained monocular camera in a deck environment. [ABSTRACT FROM AUTHOR]

To facilitate the monitoring and management of modern transportation systems, monocular visual traffic surveillance systems have been widely adopted for speed measurement, accident detection, and accident prediction. Thanks to the recent innovations in computer vision and deep learning research, the performance of visual traffic surveillance systems has been significantly improved. However, despite this success, there is a lack of survey papers that systematically review these new methods. Therefore, we conduct a systematic review of relevant studies to fill this gap and provide guidance to future studies. This paper is structured along the visual information processing pipeline that includes object detection, object tracking, and camera calibration. Moreover, we also include important applications of visual traffic surveillance systems, such as speed measurement, behavior learning, accident detection and prediction. Finally, future research directions of visual traffic surveillance systems are outlined. [ABSTRACT FROM AUTHOR]

In this work, the travel path of a set of drones is scheduled across a graph, where the nodes need to be visited multiple times at pre-defined points in time. The nodes can either be demand nodes requesting monitoring, or supply nodes that are used as take-off/landing locations for the drones and for battery replacement to cope with the limited flying range of the drones. This is an extension of the well-known multiple traveling salesman problem and the proposed formulation can be applied in several domains such as the monitoring of traffic flows in a transportation network, the monitoring of remote locations to assist search and rescue missions, or the monitoring of critical infrastructure facilities for security and surveillance purposes. Aiming to find the optimal schedule, the problem is initially formulated as an Integer Linear Program (ILP). However, given that the problem is highly combinatorial, the optimal solution scales only for small-size problems. Thus, a greedy algorithm is also proposed that uses a one-step look-ahead heuristic search mechanism, as well as an algorithm that is based on ant colony optimization (ACO). In a detailed evaluation, it is observed that both algorithms achieve near-optimal performance for small settings, while also scaling to larger settings, with the ACO being more suitable for medium-size settings and the Greedy for larger ones. A field experiment is additionally performed to demonstrate the practical implementation of the proposed system under real-world conditions. [ABSTRACT FROM AUTHOR]

With the emergence of various convolutional neural network (CNN)-based applications and the rapid growth of CNN model scale, the resource-constricted end devices can hardly deploy CNN-based applications. Current work optimizes the CNN model on edge servers and deploys the optimized model on devices in an edge computing environment. However, most of them only optimize the resource consumption within or across models solely, whereas neglecting the other side. In this article, we propose a novel CNN-based resource optimization approach (CroApp) that not only optimizes the resource consumption within the CNN model but also pays attention to resource optimization across the applications. Specifically, we adopt model compression as the “inner-model” optimization method, as well as computation sharing as the “intermodel” optimization method. First, during “inner-model” optimization, the CroApp prunes unnecessary parameters within the model on edge servers to reduce the scale of the model. Then, during “intermodel” optimization, the CroApp trains a set of shareable models based on the pruned model and sends these shareable models to end devices. Finally, the CroApp adaptively adjusts the shared models to reduce resource consumption. The experimental results show that the CroApp outperforms the state-of-the-art approaches in terms of resource reduction, scalability, and application performance. [ABSTRACT FROM AUTHOR]

This article derives a novel unified signal processing scheme for WiFi-based passive radar in order to limit its complexity and enhance its suitability for short range civilian applications. To this purpose, the exploitation of a reciprocal filtering (RpF) strategy is investigated as an alternative to conventional matched filtering at the range compression stage. Along with the well-known advantage of a remarkable sidelobes control capability for the resulting range-Doppler response, the use of a RpF is shown to provide additional benefits for the specific sensor subject of this article. Specifically, it allows to streamline the disturbance cancellation stage and to implement a unified signal processing architecture which is capable to handle the different modulation schemes typically adopted in WiFi transmissions. Appropriate adjustments are also proposed to the theoretical RpF in order to cope with the inherent loss in terms of signal-to-noise power ratio. The effectiveness of the proposed signal processing scheme encompassing the RpF strategy is proved against both simulated and experimental datasets. [ABSTRACT FROM AUTHOR]