Search

The existing open-source infrared dim and small target dataset have limited data capacity and insufficient scenarios. To improve the generalization and robustness of the network, it is usually needed to establish a new dataset for the specific application scenario. When using the conventional training method in the new dataset, there are problems such as low target-to-background proportion in full-sized images, inaccurate labels due to nonexpert annotators, and inadequate diversity of data, which affect the training efficiency and network performance. This article proposes a training strategy based on the newly established dataset to overcome these problems. Specifically, small-sized image transfer learning is proposed to increase the small target proportion, shorten training time, and improve training efficiency. Moreover, a label refinement method based on loss evaluation metrics is adopted to reduce the impact of inaccurate labeling on network training. In addition, an iterative training method is proposed by supplementing new false alarm and miss detection data into the dataset between each iteration to further improve the training performance. The experiments are carried out and the results show that the above method can effectively shorten training time, improve training efficiency and performance of infrared dim and small target detection networks.

The energy limit would abbreviate the lifetime of sweep coverage network. Wireless charging technology might be the best way to extend the lifetime of sweep coverage networks, as sensors are required to constantly move to ensure periodic target coverage. This paper studies the sweep coverage facilitated by rechargeable sensors, which persistently navigate between target nodes and charging stations to fulfill the periodic coverage demands of the targets while ensuring they return to the charging stations before their energy runs out. This paper firstly proposes a general definition of the Chargeable Sweep Coverage (CSC) problem, and studies the CSC problem by examining various constraints. We also propose one CSC problem subject to a special constraint where the sensors must return to their initial charging stations to recharge. The CSC problem under constrain is also NP-hard. In this paper, the problem is modeled as Maximum nodes inclusion problem with unbound candidate paths. We discuss the effective paths of the problems by situation, bound the candidate paths to certain lengths using discretization technology, and employ existing approximation algorithms to get the approximation solutions. When the charging period is greater than the sweep period, we get a $\frac {2}{5}$ -approximation algorithm. For the opposite, we get $\frac {1}{6}$ -approximation algorithms for it. The validity and scalability of the proposed algorithms are proved by theoretical proofs and experimental simulations.

Although the underwater optical communication system (UWOC) plays an important role in many marine applications, the acquisition, pointing and tracking (APT) problems are still great challenges in a long or turbid water channel. The Orthogonal Frequency Division Multiplexing (OFDM) and the space-time block coding (STBC) technologies have been introduced in UWOC systems to improve spectral efficiency and data transmission reliability. On these foundations, a multi-input multi-output (MIMO) mode is proposed and verified in this article to reduce the alignment requirement. In experiments, two sets of 450 nm blue laser diodes (LDs) and PIN photodiodes are selected as light sources and receivers, and then parallel light beams are generated and modulated using an Alamouti coding for its anti-interference ability. The receiving plane is rotated to simulate the various alignment situations, and the system performance of MIMO-OFDM and single-input single-output OFDM (SISO-OFDM) are compared in tap water and turbid water. The results show that the spatial diversity in MIMO can extend the communication range and its single-sided maximum detection angle range is 92% and 112% higher than the SISO system, respectively.

Signal-to-noise ratio (SNR) is an important radiation characteristic parameter for remote sensing image quality assessment as well as a key performance indicator for remote sensing sensors. At present, the SNR estimation methods based on regular segmentation or continuous segmentation are generally used to obtain image SNR. However, the land cover type has a great influence on the results of the SNR estimation method using regular segmentation, especially the high spectral resolution and high spatial resolution remote sensing images obtained by the low-altitude UAV hyperspectral sensor. In addition, some land cover types are difficult to achieve continuous segmentation. In view of this limitation of the existing SNR estimation methods, a new unsupervised method for estimating SNR in UAV hyperspectral images has been developed in this article, called pure pixel extraction and spectral decorrelation. By directly extracting pure pixels in the spatial dimension and combining the correlation of the spectral dimension to obtain the SNR of the hyperspectral image, this new method replaces the conventional method of improving the segmentation algorithm to improve the accuracy of SNR estimation. Additionally, the box counting method is introduced to determine the image SNR aggregation interval. The results showed that the proposed method had higher accuracy and smaller errors than the other SNR estimation methods. Besides, this method had stronger robustness, it can be used for both radiance and reflectance (atmospherically corrected) UAV hyperspectral images.

An AlGaInAs multiple-quantum-well (MQW) polarization mode controller (PMC) using a stepped height ridge waveguide is presented, which is monolithically integrated with a sidewall grating distributed feedback laser using quantum well intermixing (QWI). QWI is used to create a 100 nm blueshift in the PMC and to partially eliminate the anisotropy and birefringence of the MQW structure. The PMC structure is modelled and optimized using a 3D full-vectorial Finite-Element Method package. The maximum polarization conversion efficiency (PCE) is around 96% for a 537-μm-long PMC operating at a wavelength of 1550 nm. To maintain a PCE of ≥90%, the fabrication tolerances of the dry-etch corner and ridge waveguide widths are ± 0.05 μm and ± 0.03 μm respectively. The main advantages of the proposed design are that only a single step of MOPVE and two steps of dry etching are required for the whole integrated device, significantly reducing complexity and cost.

The polarization states of scattered photons can be used to map or image the anisotropic features of a nanostructure. However, the scattering strength depends heavily on the refractivity contrast in the near field under measurement, which limits the imaging sensitivity for viral particles which have little refractivity contrast with their nano-ambientes. In this paper, we show the photon scattering signal strength can be magnified by introducing a more abrupt change of refractivity at the virus particle using antibody-conjugated gold nanoparticles (AuNPs), allowing the presence of such viruses to be detected. Using two different deep learning methods to minimize scattering noise, the photon states scattering signal of a AuNPs ligated virus is enhanced significantly compared to that of a bare virus particle. This is confirmed by Finite Difference Time Domain (FDTD) numerical simulations. The sensitivity of the polarization state scattering spectra from a virus-gold particle doublet is 5.4 times higher than that of a conventional microscope image.

Pyroelectric infrared (PIR) sensors are much promising for device-free localization (DFL) due to their advantages of lower cost, low power consumption, and privacy protection. Most PIR-based localization methods usually assume some geometric models according to the detection principle of PIR sensors, which are however not accurate or robust due to the various cases of infrared radiation from human body, especially the case of multiple persons. Recently, deep learning is utilized in the PIR-based localization method (i.e. PIRNet Yang et al.) and well handles the complex infrared radiation even in the multi-person case. However, this method requires a high training cost, and has very weak generalization ability as it assumes the PIR sensors’ deployment in the testing environment is same to the deployment in training environment. To reduce the training cost and achieve high generalization ability, in this paper, we propose a robust method DeepPIRATES, which can be directly utilized in various deployment scenarios without retraining. DeepPIRATES combines deep learning and a geometric model. Specifically, DeepPIRATES divides the localization task into two steps. The first step utilizes a neural network to estimate the azimuth changes of multiple persons to a PIR sensor. Then, DeepPIRATES utilizes the persons’ azimuth changes to infer their locations based on a geometric model. Extensive experimental results show that DeepPIRATES can achieve similar localization accuracy as PIRNet, while does not require to be retrained when the sensor deployment changes.

The development of integrated electricity and heat system (IEHS) promotes the renewable energy penetration and the complementary coupling between the electric power system (EPS) and the district heat system (DHS). The efficient energy flow calculation (EFC) in the IEHS has become a hotspot as the system scale and complexity increase. This paper firstly constructed the dynamic DHS model using Euler difference scheme. On this basis, the fast decomposed method is proposed to decompose the system with various scales and topologies, which improves the computation efficiency. Moreover, the modification mode is then developed to modify the EFC results in the decomposed systems and ensure the accuracy. Numerical simulation verifies the accuracy of the proposed method and its superiority in computation efficiency.

Commanders are aware of the battlefield situation (BS) by monitoring the battlefield periodically. When the monitoring interval period is too short, there will be a huge cognitive load, which seriously affects commanders’ situation awareness (SA). When the monitoring interval period is too long, there will be omission of important information, which also affects commanders’ SA. Therefore, it is necessary to set the reasonable interval period of monitoring battlefield. In this paper, BS signal is defined to represent BS changes, according to the characteristics of BS changes. Then Fourier transform (FT) is used to analyze frequency characteristics of BS signal in order to describe the complicatedness of BS changes. The discriminant of reasonable monitoring interval period (DRMIP) is derived, according to the commanders’ requirements accuracy of BS. When the complicatedness of BS changes cannot be obtained previously, the highest frequency iteration algorithm (HFIA) is proposed to obtain the best monitoring interval period according to iterative BS signal. In the case of setting the threshold of difference, the validity of the proposed method was verified by simulation calculation method.

Locating the tiny insulator defect object with complex backgrounds in high-resolution aerial images is a challenging task. In this paper, we propose a novel method which cascades detection and segmentation networks to identify the defect from the global and local two levels: (1) The improved Faster R-CNN is carried out to capture both defects and insulators in the entire image. ResNeXt-101 is adopted as the feature extraction network so as to fully extract features, and Feature Pyramid Network (FPN) is built to enhance the ability of detecting small targets. In addition, the Online Hard Example Mining (OHEM) training strategy is applied to solve the imbalance problem of positive and negative samples. (2) All the detected insulators are extracted and fed into the improved U-Net network to futher inspect at pixel level, we utilize the pre-trained ResNeXt-50 as the encoder of U-Net, incorporate an attention module, Spatial and Channel Squeeze & Excitation Block (SCSE), into the decoding path to highlight the meaningful information. A hybrid loss which merges binary cross entropy (BCE) loss and dice coefficient loss is designed to train our network for figuring out the class imbalance issue. The missed detection can be greatly reduced with the combination of two modified network, which makes comprehensive use of the original map information and local information. On the test set of actual images, the insulator defect recognition precision and recall of the cascade network is 91.9% and 95.7%, exhibiting strong robustness and accuracy.

Split air conditioner (SAC) plays a critical role for temperature control of buildings. In recent years, the manufacturing process of SAC tends to consider energy-saving and environmental protection factors to reduce air pollution and alleviate energy crisis. However, the design of environmental-friendly SAC will increase the manufacturing cost, thereby increasing the burden on air-conditioning companies. Therefore, the eco-cost analysis of SAC is of great significance for air conditioner manufacturers. To this end, a new SAC eco-cost assessment method based on activity based costing method is innovatively proposed in this paper. In this method, a K-nearest neighbor algorithm is used to deal with the missing data in SAC samples. An environmental and non-environmental cost assessment approach is developed based on the activity-based costing method to estimate the eco-cost of SAC over its whole life cycle. In addition, the impact of energy efficiency deviation on SAC eco-cost is numerically analyzed. Finally, the proposed eco-cost assessment method is programmed on a Gabi software, and validated using real SAC data from a Chinese air-conditioning company. Simulation results show that the proposed method can provide an estimate of the eco-cost of SAC, thus helping air-conditioning companies to improve their competitiveness in the future where energy-saving and emission reduction are increasingly important.

Accurate and reliable prediction of air-conditioning load plays a significant role in prosumer energy management system (EMS), because air-conditioning load accounts for a large proportion of the building's total energy consumption. This paper proposes a new meta ensemble learning method to realize short-term prediction of air-conditioning load for prosumers. This method is a hybrid of meta ensemble learning and stacked auto-encoder (SAE). First, we design multiple different forecasting structures based on SAE to achieve point prediction of air-conditioning loads. SAE is used to learn the deep features in air-conditioning load data. Second, a new meta ensemble learning prediction model is proposed. Meta ensemble learning is used to learn the nonlinear features and invariant structures in data, and determine the coefficients of each SAE-based point forecaster. Finally, the prediction results of each point forecaster are aggregated and integrated to estimate the final air-conditioning load prediction result. Air-conditioning load data from a commercial building in Singapore are used to validate the feasibility and effectiveness of the proposed method, demonstrating that the proposed meta ensemble learning method is attractive in prosumer energy management.

Because social networks exemplify the phenomenon of homogeneity in complex networks, researchers generally believe that a user's privacy disclosure is closely related to that of the users around them, but we find that the related users studied in previous methods were not correct. That is, the analyzed user groups may have had nothing to do with the privacy disclosure of the target users. Since private information is time-sensitive, information held by users who are no longer in the same environment as the target user may no longer be true and have lost its value. For example, considering students and members of the working class, transfers to another school for further studies or job changes entail dramatic changes to most of their information. This lack of timeliness has an overarching impact on the effectiveness of social network analysis and privacy protection, but this problem has not been addressed by researchers. Therefore, we study and characterize this problem, add the user's behavior trace to solve this problem and measure the user's privacy status more accurately.

Energy consumption predictions for residential buildings play an important role in the energy management and control system, as the supply and demand of energy experience dynamic and seasonal changes. In this paper, monthly electricity consumption ratings are precisely classified based on open data in an entire region, which includes over 16 000 residential buildings. First, data mining techniques are used to discover and summarize the electricity usage patterns hidden in the data. Second, the particle swarm optimization-K-means algorithm is applied to the clustering analysis, and the level of electricity usage is divided by the cluster centers. Finally, an efficient classification model using a support vector machine as the basic optimization framework is proposed, and its feasibility is verified. The results illustrate that the accuracy and F-measure of the new model reach 96.8% and 97.4%, respectively, which vastly exceed those of conventional methods. To the best of our knowledge, the research on predicting the electricity consumption ratings of residential buildings in an entire region has not been publicly released. The method proposed in this paper would assist the power sector in grasping the dynamic behavior of residential electricity for supply and demand management strategies and provide a decision-making reference for the rational allocation of the power supply, which will be valuable in improving the overall power grid quality.

Emerging research revealed that the view-type of photos is not only related to the field of data sciences, such as the sentiment brought forth by sightseeing spots, but also in the field of social sciences about human emotions and behaviors. These potential usages of view-types trigger a challenging problem, that is to automatically distinguish them into wide or narrow. In this paper, we present a computational model to classify them inspired by the human visual system. We found two cues that can represent the visual attention, i.e., focus cue and scale cue. The focus cue is modeled in the frequency domain using the non-sampled contourlet transform (NSCT) and speeded up robust features (SURF). The scale cue is modeled by defining the spatial size and conceptual sizes of an object in the image, whereby AdobeBING and convolutional neural network are used for the respective measurements. By integrating these focus and scale models, a robust scheme is hence proposed for this non-trivial task. The experiments on a newly established dataset, which has 5050 natural images, show better performance by our proposal when compared to the state-of-the-arts.

The face recognition has become a common means of identity authentication because of the advantages of uniqueness, non-invasive and not easy to be stolen. The outsourcing of face recognition to the service provider is a typical manner nowadays. However, it raises vital concerns about the privacy of outsourcing server due to the sensitivity of face data. Therefore, a frame of identity authentication based on the technology of privacy-preserving face recognition is presented in this paper. The convolutional neural network is used for face feature extraction. To overcome the issue of privacy leaked, a method of secure nearest neighbor that can compute the cosine similarity over encrypted feature vectors is proposed. What's more, the edge computing is introduced in our frame to increase the authentication efficiency by removing some operations from the cloud to the edge of the Internet. Moreover, we also propose a secret sharing homomorphism technology which is used for distributed computing to improve the fault-tolerance of our identity authentication system. The experimental results show that the proposed schemes are secure and effective.

In recent years, artificial intelligence has played an increasingly important role in the field of automated control of drones. After AlphaGo used Intensive Learning to defeat the World Go Championship, intensive learning gained widespread attention. However, most of the existing reinforcement learning is applied in games with only two or three moving directions. This paper proves that deep reinforcement learning can be successfully applied to an ancient puzzle game Nokia Snake after further processing. A game with four directions of movement. Through deep intensive learning and training, the Snake (or self-learning Snake) learns to find the target path autonomously, and the average score on the Snake Game exceeds the average score on human level. This kind of Snake algorithm that can find the target path autonomously has broad prospects in the industrial field, such as: UAV oil and gas field inspection, Use drones to search for and rescue injured people after a complex disaster. As we all know, post-disaster relief requires careful staffing and material dispatch. There are many factors that need to be considered in the artificial planning of disaster relief. Therefore, we want to design a drone that can search and rescue personnel and dispatch materials. Current drones are quite mature in terms of automation control, but current drones require manual control. Therefore, the Snake algorithm proposed here to be able to find the target path autonomously is an attempt and key technology in the design of autonomous search and rescue personnel and material dispatching drones.

Experiments used for model validation constitute a new type of experiments, whose primary goal is to determine the ability of a high-fidelity mathematical model in simulating a well-characterized physical process. Considering the importance of the credibility of the validation experiment, in this paper, a methodology for the optimal design of validation experiments based on the area metric factor and the fuzzy expert system is presented to obtain the test scheme with low cost and satisfactory credibility. First, the concept of area metric factor is proposed, which is a dimensionless validation metric. Then, a criterion for qualitative assessment of the predictive model accuracy is developed based on the fuzzy expert system, in which the inconsistency among the different expert groups is considered. Subsequently, an optimization model of the validation experiment design is constructed, in which the sample size of the experiments is designed variable, while the credibility of the validation experiment is a constraint. Meanwhile, a novel method is developed to solve the optimization model, in which the Latin hypercube sampling is used to obtain the location of each validation experiment. Finally, two simulation examples are adopted to validate the proposed methods. The simulation results show that the randomness of the experimental scheme significantly affects the credibility of the evaluation results in the case of small sample sizes. The proposed method for the validation experiment design can reduce the impact of the selection of the testing programs on the credibility of model validation.

Target detection and classification in the shadow area of hyperspectral images (HSIs) has always been an important problem in the field of HSI data processing. However, there are few methods to detect or classify targets in the shadow area effectively because of occlusion of objects or oblique solar radiation. Dynamic stochastic resonance (DSR) theory shows that under the synergistic action of weak input signal, noise and non-linear system, the energy of noise can be transferred into signal partially. It breaks the idea that signal can be enhanced only by eliminating noise and has been proved to be effective in many fields. In this paper, DSR is introduced into the shadow area enhancement of HSI from both spatial and spectral dimensions. Then, the enhanced HSI data can be obtained by fusing the processed shadow with the original HSI data. Finally, 2D convolutional neural networks (2D-CNN) is used to classify the enhanced HSI. To evaluate the proposed method, a real-world HSI has been processed in the experiment and the results show that DSR can improve the contrast and the spectral intensity of HSI shadow area. Compared to other state-of-the-art methods, the classification accuracy is better at containing different targets with small samples, especially in the spectral dimension.

With the development of electric vehicles in recent years, lithium-ion batteries have been widely used. Accurate state of charge (SOC) estimation plays an important role in the safety of electric vehicles. Since the temperature has the significant influence on charge and discharge performance of the battery, it is critical to achieve accurate SOC estimation over the wide temperature range. In this paper, a polymer ternary lithium-ion battery is focused, and a Thevenin equivalent circuit model with temperature compensation is established. The validity of the established battery model was verified by the dynamic stress test. On this basis, the ternary lithium-ion battery SOC was estimated using the unscented Kalman filter (UKF). The New European Driving Cycle is used to verify the effectiveness of the proposed algorithm. The simulation and experimental results show that the established Thevenin equivalent circuit model with temperature compensation can accurately represent the battery dynamics. Based on this model, the SOC was estimated using the UKF and the maximum errors are within 3%. Therefore, the proposed SOC estimation method is verified to be effective and robust.

Using multiple fixed chargers to remotely charge wireless sensor networks (WSNs) is a feasible way in harsh terrain. An interesting property is that, due to the radio interference, the charging efficiency with multiple chargers may not simply be the sum of that with single charger if the chargers have the same radio frequency. To avoid the interference, an alternative way is to employ chargers with diverse frequencies, which, however, occupies too much channel resources of WSNs. In this paper, we try to answer such an interesting question: to charge a given WSN, can unique-frequency chargers be comparable to diverse-frequency chargers? To answer this question, we formulate two problems targeting at minimizing the charger number and maximizing the minimum charging power at sensor nodes, respectively. Then, we propose corresponding greedy algorithms with proof of the approximation ratio. Extensive simulation results show that: 1) given threshold of radio power at sensor nodes, the number of unique-frequency chargers required is no more than 1.1 times of that of diverse-frequency chargers and 2) given the charger number, the minimum radio power provided by unique-frequency chargers can be up to 80% of that with diversefrequency chargers. This shows the quite competitive performance of unique-frequency chargers since they require only one channel instead of a group of channels required by diverse-frequency chargers. In addition, after appropriate placement, we find it needless to further schedule the unique-frequency chargers to improve the minimum charging power that they provide.

In online social networks (OSNs), high trust value entities play an important role in service recommendation when users inquire certain service. Generally, users in OSNs are more willing to choose those services recommended by high trust value entities. In fact, users may suffer from great loss of property once they accept some bad services provided by high trust value entities. However, current schemes do not consider this problem. Hence, we propose a scheme called RHT (recommendation from high trust value entities) to evaluate the trust degree of service recommended by high trust value entities. To be specific, there exist other users who provide their ratings to the service recommended by a high trust value entity, and RHT first selects the trusted ones from those users by computing the similarity between target user and them. Simultaneously, RHT also withstands malicious attacks during the trusted nodes selection. In addition, we also design an adaptive trust computation method to calculate trust value according to the ratings of trusted users. The experimental results show that RHT has higher accuracy in trust evaluation compared with current representative schemes and do effectively resistant four common attacks when choosing trusted nodes.

Hybrid precoding has been regarded as a promising technology for millimeter-wave (mmWave) massive multiple-input multiple-output (MIMO) systems due to the low hardware cost and power consumption. This paper focuses on digital and analog codebooks design to efficiently quantize the hybrid precoders with arbitrary-resolution analog phase shifters in frequency-division duplex systems. We formulate the codebooks design as a joint optimization problem and propose an iteration algorithm to obtain the enhanced codebooks based on dictionary learning methods. Moreover, this paper shows that the optimal precoders for all data streams exhibit a hidden joint sparsity structure caused by the limited number of scattering clusters with low angular spreads of mmWave MIMO channels. Exploiting the structure, a jointly sparse precoding is proposed as the pursuit method for the proposed codebooks design algorithm. The simulation results demonstrate that the near-optimal performance can be achieved by the joint utilization of the proposed hybrid precoding algorithm and codebooks even when the low-resolution phase shifters are used.

The cutting efficiency in underground excavations relies on the optimum parameters of the cutting tool and the cutting process. However, the optimization of the cutting tool design and the cutting process is a challenge and requires knowledge about the tool-rock interaction. This paper aims to investigate the tool-rock interaction using a rock cutting mathematical model. The confining pressure was considered in the rock cutting model with conical cutters and the discrete element method was adopted to calculate the dynamics of the rock breakage of this model. Graded particle assemblies were created, calibrated, and compressed in the horizontal direction with a certain confining pressure. Afterwards, the initiation and propagation of cracks during the rock cutting processes were recorded. A series of small-scale rock cutting tests were also carried out to verify the numerical model. The analysis results demonstrate that: 1) the confining pressure induced larger cutting force than that in the unconfined condition; 2) with increase of the confining pressure, the rock failure mode experienced predominantly brittle to predominantly ductile failure; and 3) there was a critical confining pressure/compressive strength ratio of 0.53 when the transition of failure mode occurred.

In this paper, we report a novel fiber-optic Fabry-Perot interferometric (FFPI) temperature sensor based on the difference of thermal expansion coefficient between fused silica and metallic materials. The sensor head is made by a single-mode fiber (SMF). A gold film and a nickel film are sputtered and electroplated on the surface of the SMF. Then, a microcavity is micromachined by focused ion beam (FIB) milling. Because the thermal expansion coefficient of nickel is about 20 times of fused silica, the different thermal expansions force the sensor head to bend when the temperature is high or low. Its temperature sensitivity is over 14 pm/°C in a wide range from -79°C to +70°C. And the coefficient of determination R2 is excellent (over 0.995). Moreover, the metallic cylinder can reinforce the cavity spot of the fiber sensor, so that this kind of sensor can work in harsh environments. For the first time to the best of our knowledge, we report this type of FFPI temperature sensor based on difference of thermal expansion coefficient between fused silica and metallic materials.