Search

Surveillance videos are essential for crime prevention and public safety. However, defining abnormal events remains challenging, which hinders their effectiveness and limits the use of supervised techniques. The existing method have difficulty in accurately detecting and tracking the objects, that minimizes the detection and tracking performance. In this research, the You Only Look Once v7 (YOLO v7) - Distance Intersection of Union (D-IoU) and Earth Mover Distance (EMD) approach is proposed to detect and track the objects and anomalies in video surveillance. The D-IoU loss function is used in the YOLO v7 model improves the precision of bounding box prediction by considering the distance between centres of the bounding box, which is useful for the accurate localization of object. Then, the features are extracted by using the Inception V3 approach that extracts the meaningful features that help differentiate the anomalies in the detected object. To detect the anomalies, the EMD method is used which effectively detects the anomalies. The YOLO v7 – D-IoU and EMD approach obtained 96.1% accuracy on UCSD Ped 1 datasets and 98.8% accuracy on UCSD Ped 2 dataset. The proposed method showed effective performance when compared to conventional methods like Three-Dimensional Convolutional Neural Network (3D-CNN). [ABSTRACT FROM AUTHOR]

The development of rapid safety assessment tools for signalized intersections is a pivotal focus in contemporary research. Recent advancements in technology and artificial intelligence (Al) have driven the development of proactive safety assessment tools that highlight hazardous instances and provide accurate safety insights. These tools utilize various detection methods, such as Closed-Circuit Television (CCTV) cameras, Light Detection and Ranging (LiDAR), Unmanned Aerial Vehicles (UAVs), infrared (IR), and loop detectors. For nationwide deployment, the most cost-effective methods use existing surveillance cameras. This study investigates different computer vision detection algorithms, focusing on training datasets, video analysis duration, detection methods, types of detected objects, and strengths and weaknesses of each algorithm. It examines convolutional neural network (CNN) detection algorithms based on bounding boxes, 3-D bounding boxes, and key points. These algorithms were applied to video footage from two case study intersections: the Town Square intersection in Jackson Hole, Wyoming, with three cameras in a rural setting, and the Four Comer Camera intersection in Cold Water, Michigan, with one fixed camera in an urban environment. Key findings suggest that 3-D bounding box detection algorithms are more effective at lower elevations for estimating occluded parts and extracting accurate trajectories. Higher-elevation cameras benefit more from bounding box algorithms for faster processing, while key point detection algorithms excel at intersections with multiple cameras, providing accurate depictions and localization of road users. These results offer valuable recommendations for developing accurate, cost-effective, and time-efficient safety assessment tools. [ABSTRACT FROM AUTHOR]

Autonomous driving (AD) technology has seen significant advancements in recent years; however, challenges remain, particularly in achieving reliable performance under adverse weather conditions such as heavy fog. In response, we propose a multi-class fog density classification approach to enhance the AD system performance. By categorizing fog density into multiple levels (25%, 50%, 75%, and 100%) and generating separate datasets for each class using the CARLA simulator, we improve the perception accuracy for each specific fog density level and analyze the effects of varying fog intensities. This targeted approach offers benefits such as improved object detection, specialized training for each fog class, and increased generalizability. Our results demonstrate enhanced perception of various objects, including cars, buses, trucks, vans, pedestrians, and traffic lights, across all fog densities. This multi-class fog density method is a promising advancement toward achieving reliable AD performance in challenging weather, improving both the precision and recall of object detection algorithms under diverse fog conditions. [ABSTRACT FROM AUTHOR]

To address the problem of managing dedicated parking zones arising from the increasing number of electric vehicles and vehicles for the physically challenged, this paper proposes a license plate recognition (LPR)-based parking control system that combines the YOLO and MobileNet algorithms. These two algorithms are designed for real-time object detection and efficient preprocessing, respectively, and can operate in real time in resource-constrained edge-device environments. In tests using data from more than 51,000 vehicles, the system achieved an accuracy rate of 95.76% in classifying electric vehicles and 97.18% in classifying vehicles for the physically challenged. The average CPU and RAM utilizations of the system were 34.54% and 45.04%, respectively. In addition, the processing time per image was recorded as approximately 1.04 s, demonstrating its potential to run reliably on edge devices. These results are expected to facilitate the efficient resolution of parking management problems in smart cities and effective operation of parking zones reserved for electric vehicles and vehicles for the physically challenged.

Abstract Loading water drilling rig on the anchor digging machine can effectively increase the tunneling efficiency. In order to avoid the interference between the water drilling rig and the anchor machine in the working process, it is necessary to calculate the joint variables of the drilling rig accurately. Using the robot kinematics analysis method, the kinematics model of the system is established. At the same time, combined with bounding box collision detection technology, the key parts of the equipment are mathematically modeled. A method for collision detection of onboard water exploration and drainage drilling rig of anchor digging machine is proposed. The simulation results show : The designed collision detection method can be used to guide the attitude adjustment of the equipment during the drilling operation of the equipment, and can accurately analyze the interference characteristics and collision detection of the equipment. It provides a theoretical basis for the automatic excavation of the research object and the automatic control of drilling. The azimuth adjustment range of the water drilling rig for the experiment is : ± 15°, and the pitch angle adjustment range is : + 5° to -10°.

Autonomous driving (AD) technology has seen significant advancements in recent years; however, challenges remain, particularly in achieving reliable performance under adverse weather conditions such as heavy fog. In response, we propose a multi-class fog density classification approach to enhance the AD system performance. By categorizing fog density into multiple levels (25%, 50%, 75%, and 100%) and generating separate datasets for each class using the CARLA simulator, we improve the perception accuracy for each specific fog density level and analyze the effects of varying fog intensities. This targeted approach offers benefits such as improved object detection, specialized training for each fog class, and increased generalizability. Our results demonstrate enhanced perception of various objects, including cars, buses, trucks, vans, pedestrians, and traffic lights, across all fog densities. This multi-class fog density method is a promising advancement toward achieving reliable AD performance in challenging weather, improving both the precision and recall of object detection algorithms under diverse fog conditions.

A Minimum Bounding Box (MBB) is a rectangle which bounds a geographic feature or dataset. It is commonly used in spatial information systems as a simplified way of describing the spatial extent of a resource. MBBs are typically indexed for searching and discovering resources relevant to a given geographic area of interest. However, this simplification leads to a loss of precision in the description of the extent and can affect the overall precision of the search results. We propose an alternative technique for describing the spatial extent based on the use of DGGS tiles. To measure the precision improvements offered by our method, we designed and implemented an empirical method for evaluating the average precision, and applied it to three different systems: one based on MBB, another on Convex Hull, and ours based on DGGS. The three methods were evaluated with the same test collection obtained from some of the main European geospatial data catalogues compliant with the INSPIRE directive. The results showed that our method outperformed the other two. Where the catalogue average precision of the MBB search scenarios is between 73% and 97%, the DGGS is between 96% and 99%. Additionally, we propose a realistic method of transitioning from the current technologies to the technology we are proposing, considering the current state of the spatial data infrastructures. [ABSTRACT FROM AUTHOR]

Connected cooperative and automated (CAM) vehicles and self-driving cars need to achieve robust and accurate environment understanding. With this aim, they are usually equipped with sensors and adopt multiple sensing strategies, also fused among them to exploit their complementary properties. In recent years, artificial intelligence such as machine learning- and deep learning-based approaches have been applied for object and pedestrian detection and prediction reliability quantification. This paper proposes a procedure based on the YOLOv8 (You Only Look Once) method to discover objects on the roads such as cars, traffic lights, pedestrians and street signs in foggy weather conditions. In particular, YOLOv8 is a recent release of YOLO, a popular neural network model used for object detection and image classification. The obtained model is applied to a dataset including about 4000 foggy road images and the object detection accuracy is improved by changing hyperparameters such as epochs, batch size and augmentation methods. To achieve good accuracy and few errors in detecting objects in the images, the hyperparameters are optimized by four different methods, and different metrics are considered, namely accuracy factor, precision, recall, precision–recall and loss. [ABSTRACT FROM AUTHOR]

The development of deep learning technology is widely used for various purposes, including recognizing characters in a document. One of the scripts that can benefit from this deep learning technology is the Komering script, which is a local script in the South Sumatra region. However, there are challenges in reading documents written in this script, requiring a method to separate each character in a document. Therefore, there is a need for a technology that can automatically segment images of documents written in the Komering script. This research introduces an innovative technique for segmenting images of characters in documents that contain Komering script characters. The segmentation technique employs bounding box technology to separate each Komering script character, subsequently recognized by a pre-trained deep learning model. The bounding box approach imposes restrictions on the segmented object area. To recognize Komering characters, a deep learning model with a convolutional neural network (CNN) algorithm is employed. [ABSTRACT FROM AUTHOR]

The increasing rate of minimally invasive procedures and the growing prevalence of cardiovascular disease have led to a demand for higher-quality guidance systems for catheter tracking. Traditional methods for catheter tracking, such as detection based on single points and applying masking techniques, have been limited in their ability to provide accurate pose information. In this paper, we propose a novel deep learning-based method for catheter tracking and pose detection. Our method uses a Yolov5 bounding box neural network with postprocessing to perform landmark detection in four regions of the catheter: the tip, radio-opaque marker, bend, and entry point. This allows us to track the catheter’s position and orientation in real time, without the need for additional masking or segmentation techniques. We evaluated our method on a dataset of fluoroscopic images from two distinct datasets and achieved state-of-the-art results in terms of accuracy and robustness. Our model was able to detect all four landmark features (tip, marker, bend, and entry) used to generate a pose for a catheter with 0.285 ± 0.143 mm, 0.261 ± 0.138 mm, 0.424 ± 0.361 mm, and 0.235 ± 0.085 mm accuracy. We believe that our method has the potential to significantly improve the accuracy and efficiency of catheter tracking in medical procedures that utilize bi-plane fluoroscopy guidance.

The increasing rate of minimally invasive procedures and the growing prevalence of cardiovascular disease have led to a demand for higher-quality guidance systems for catheter tracking. Traditional methods for catheter tracking, such as detection based on single points and applying masking techniques, have been limited in their ability to provide accurate pose information. In this paper, we propose a novel deep learning-based method for catheter tracking and pose detection. Our method uses a Yolov5 bounding box neural network with postprocessing to perform landmark detection in four regions of the catheter: the tip, radio-opaque marker, bend, and entry point. This allows us to track the catheter's position and orientation in real time, without the need for additional masking or segmentation techniques. We evaluated our method on a dataset of fluoroscopic images from two distinct datasets and achieved state-of-the-art results in terms of accuracy and robustness. Our model was able to detect all four landmark features (tip, marker, bend, and entry) used to generate a pose for a catheter with 0.285 ± 0.143 mm, 0.261 ± 0.138 mm, 0.424 ± 0.361 mm, and 0.235 ± 0.085 mm accuracy. We believe that our method has the potential to significantly improve the accuracy and efficiency of catheter tracking in medical procedures that utilize bi-plane fluoroscopy guidance. [ABSTRACT FROM AUTHOR]

The rapid increase in the number of vehicles has led to increasing traffic congestion, traffic accidents, and motor vehicle crime rates. The management of various parking lots has also become increasingly challenging. Vehicle-type recognition technology can reduce the workload of humans in vehicle management operations. Therefore, the application of image technology for vehicle-type recognition is of great significance for integrated traffic management. In this paper, an improved faster region with convolutional neural network features (Faster R-CNN) model was proposed for vehicle-type recognition. Firstly, the output features of different convolution layers were combined to improve the recognition accuracy. Then, the average precision (AP) of the recognition model was improved through the contextual features of the original image and the object bounding box optimization strategy. Finally, the comparison experiment used the vehicle image dataset of three vehicle types, including cars, sports utility vehicles (SUVs), and vans. The experimental results show that the improved recognition model can effectively identify vehicle types in the images. The AP of the three vehicle types is 83.2%, 79.2%, and 78.4%, respectively, and the mean average precision (mAP) is 1.7% higher than that of the traditional Faster R-CNN model. [ABSTRACT FROM AUTHOR]

In public spaces, conducting security checks to detect concealed objects carried on the human body is crucial for enhancing global anti-terrorist measures. Terahertz imaging has recently played a pivotal role in concealed object detection. However, previous studies have faced significant challenges in achieving superior accuracy and performance. To address these issues, we propose a YOLOv5m model for detecting hidden objects beneath human clothing. We employ the CSPDarknet53 block to reduce noise and enhance discriminative power. Object location and size are identified using a PANet and the prediction head. To reduce computational complexity and obtain highly relevant features, we utilize multi-convolutional layers. Duplicate boxes are eliminated and high-quality bounding boxes are accurately detected using the NMS block. Hyper parameter tuning is performed using the Mutation Enabled Salp Swarm Algorithm, resulting in improved detection accuracy and reduced processing time. Our proposed model achieves impressive metrics, including a precision of 98.99%, recall of 97.80%, F1 score of 98.05%, detection rate of 96.50% and execution time of 135 s. Comparatively, our method outperforms existing approaches such as CNN, YOLO3, AC-SDBSCAN, YOLO-v2, RaadNet and SPFAN. We train and test our proposed method using a terahertz video dataset, demonstrating excellent results with high precision. [ABSTRACT FROM AUTHOR]

Purpose: This study addresses the optimization process of a horizontal washing machine suspension modeled after a dynamic 7 DOF model. Methodology: A dynamical model of a horizontal washing machine is built. A simple genetic algorithm is applied to determine the optimal parameters for damping, stiffness, and position of the suspension elements by minimizing the weighted fitness function. The fitness function uses the bounding box to enclose the kinematic data of the washing machine mass center and minimize them. Finally, the Bode diagrams were used to evaluate the general aspect of the genetic algorithm performance. Results and Conclusions: The results show remarkable improvements in the displacement, velocity, and acceleration of the center of mass of the washing machine, drastically reducing vibrations and reducing the time needed to reach the final spin speed. [ABSTRACT FROM AUTHOR]

The utilization of Synthetic Aperture Radar (SAR) images for ship recognition holds significant importance within the realm of maritime surveillance and security. SAR images are useful for ship detection and recognition because they can penetrate through clouds and capture detailed information about ships, such as their size, shape, and orientation. In the context of ship recognition using SAR images, the primary objective is to employ automated methods for the identification and categorization of ships present in SAR imagery. The detection of ships in SAR images is a significant research area, but it remains challenging due to speckle noise, land clutters, and low signal-to-noise ratio. Researchers have developed various approaches to overcome this challenge, such as adaptive filtering, speckle reduction, and segmentation techniques. Hence, a ship detection method is devised that combines the active contour method and the YOLO-v8 model using deep learning techniques. In the first step, the SAR images undergo pre-processing and normalization, and the model is trained with the backbone network. The YOLO-v8 model, renowned for its proficiency in object detection, is applied to delineate precise bounding boxes around ships within the images. The results obtained from experiments conducted on a variety of SAR images convincingly demonstrate that the suggested approach attains proficient ship target detection, striking a balance between accuracy and comprehensiveness. This approach represents a promising solution to enhance ship detection in challenging SAR scenarios. [ABSTRACT FROM AUTHOR]

This study introduces a reactive trajectory generation framework designed for navigating a hexacopter through environments with multiple obstacles. The algorithm uses the obstacle data acquired by a LiDAR sensor mounted on the UAV to dynamically generate trajectories in real-time. As the UAV continuously acquires obstacle data during the flight, spherical bounding boxes are created for identifying the safe navigational spaces, thereby effectively reducing collision risks. Moreover, an interior point algorithm is employed to determine the aiming points on these bounding boxes, which optimizes the trajectory generation. Additionally, the framework incorporates a Fuzzy c-means clustering algorithm, which enables the UAV to dynamically detect and maneuver around multiple obstacles. The effectiveness and robustness of the proposed algorithm are rigorously tested through single-obstacle scenarios and extensive Monte Carlo simulations, which confirmed its viability in environments with multiple obstacles.

At present, accidents caused by falls causing loss in the elderly and those living alone. The resulting loss can lead to disability or even death if not treated in time. Therefore, the developers propose a new technique for detecting falls. It uses image processing from IP cameras with MediaPipe and bounding box technology to help fall and gesture detection. And also adds to the identification of the person who was involved in the accident, which uses LBPH technology to know who the accident victim is. As for notifications, Line Notify is used to allow program users to access notifications easily and quickly. The accuracy of the test was 93.1% for identifying faces and 90.5% for detecting falling gestures. ปัจจุบันอุบัติเหตุที่เกิดจากการหกล้ม ทำให้เกิดการสูญเสียในผู้สูงอายุและผู้ที่อาศัยอยู่คนเดียว ความสูญเสียที่เกิดขึ้นอาจทำให้ถึงขั้นพิการหรือเสียชีวิตได้หากไม่ได้รับการรักษาที่ทันเวลา ดังนั้นผู้พัฒนาจึงขอเสนอเทคนิคใหม่สำหรับการตรวจจับการหกล้ม โดยใช้การประมวลภาพจากกล้อง IP ด้วยเทคโนโลยีมีเดียไปป์และกล่องขอบเขตมาช่วยในส่วนของการตรวจจับการหกล้มและท่าทาง และยังเพิ่มเติมในส่วนของการระบุตัวผู้ที่ประสบเหตุซึ่งจะใช้เทคโนโลยี LBPH ให้รู้ได้ว่าผู้ประสบอุบัติเหตุเป็นใคร ส่วนการแจ้งเตือนจะใช้ Line Notify เพื่อให้ผู้ใช้งานโปรแกรมเข้าถึงการแจ้งเตือนได้ง่ายและสะดวกรวดเร็ว ผลการทดสอบความแม่นยำในการระบุใบหน้าร้อยละ 93.1 และตรวจจับท่าทางการล้มร้อยละ 90.5

Most plant diseases have observable symptoms, and the widely used approach to recognize plant leaf disease is by visually examining the affected plant leaves. A model that might perform the feature extraction without errors will process the classification task successfully. The technology has limitations, such as high parameters, slow detection, and inadequate performance in detecting small dense spots. These factors restrict the practical applications of technology in the field of agriculture. Hence, this work focuses on devising an optimized framework based on YOLOv7 that encompasses pre-processing and hybrid optimization techniques. This proposed YR2S (YOLO-Enhanced Rat Swarm Optimizer - Red Fox Optimization (RFO-ShuffleNetv2) has been devised. After the pre-processing, feature maps are generated using PCFAN. Later, these feature maps are used for the detection of leaves. ShuffleNet with ERSO is used to optimize the classification process. Segmentation of the area prone to disease could be identified through the FCN-RFO. This framework is deployed on the customized dataset, which comprehends images of various plant leaves. The leaf disease dataset is used for simulating and assessing the model. The experimental analysis reveals that the proposed method can effectively classify and detect leaf disease with high accuracy, i.e., 99.69%, outperforming the state-of-the-art approaches in the literature. Practical implication shows that the proposed deep learning classifiers are efficient and highly accurate.

The complexity of the environment limits the accuracy of the traditional Adaptive Monte Carlo Localization(AMCL) algorithm, which also suffers from high computational effort and particle degradation due to laser model limitations. To address these issues, an optimized AMCL algorithm with a bounding box is proposed. The AMCL algorithm is first parameterized and initialized to the particle swarm. During the particle iteration process, collision detection is performed on the bounding box. If a collision occurs, the particle filter is not updated and its particle weight is set to 1. If there is no collision, the particle filter is updated normally and the particle weight is set to 0. Then, the particles are resampled and updated based on the measurement data and motion model. After experimental verification, this method’s self-localization trajectory is closer to the actual path, and the measurement error fluctuation is smaller. The RVIZ simulation experiments revealed that the overall positioning time was optimized by 18.25% compared to the original AMCL, and by 9.28% compared to the improved AMCL. The optimization algorithm effectively improved the positioning accuracy and robustness of the system.

Indonesia boasts a diverse range of indigenous scripts, called Nusantara scripts, which encompass Bali, Batak, Bugis, Javanese, Kawi, Kerinci, Lampung, Pallava, Rejang, and Sundanese scripts. However, prevailing character detection techniques predominantly cater to Latin or Chinese scripts. In an extension of our prior work, which concentrated on the classification of script types and character recognition within Nusantara script systems, this study advances our research by integrating object detection techniques, employing the YOLOv5 model, and enhancing performance through the incorporation of the U-Net model to facilitate the pinpointing of fundamental Nusantara script's character locations within input document images. Subsequently, our investigation delves into rearranging these character positions in alignment with the distinctive styles of Nusantara scripts. Experimental results reveal YOLOv5's performance, yielding a loss rate of approximately 0.05 in character location detection. Concurrently, the U-Net model exhibits an accuracy ranging from 75% to 90% for predicting character regions. While YOLOv5 may not achieve flawless detection of all Nusantara scripts, integrating the U-Net model significantly enhances the detection rate by 1.2%.

Baby Fish counting must be counted accurately so it will not cause any loss, especially for fish seeds or fingerlings that have a small size. Generally, people still use conventional counting methods that produce low accuracy values. This research will make a Nila Baby Fish fingerlings counter program using the YOLOv3 algorithm and Blobb detection technique. The annotation data process will use LabelImg, and the dataset training will use Google COLABoratory with the Darknet framework in an online environment. Images that will predict in this program will be called and detected with an object detector. The object with a confidence score of more than 0.3 will be converted into a blob. The blob value will be forwarded to the output layer for scaling the bounding box objects. The output of this program is the predicted image, blob value, prediction time, and the number of Nila seeds. The model performance is evaluated using a confusion matrix and got a 98.87% for accuracy score.

Camera-based object tracking systems in a given closed environment lack privacy and confidentiality. In this study, light detection and ranging (LiDAR) was applied to track objects similar to the camera tracking in a closed environment, guaranteeing privacy and confidentiality. The primary objective was to demonstrate the efficacy of the proposed technique through carefully designed experiments conducted using two scenarios. In Scenario I, the study illustrates the capability of the proposed technique to detect the locations of multiple objects positioned on a flat surface, achieved by analyzing LiDAR data collected from several locations within the closed environment. Scenario II demonstrates the effectiveness of the proposed technique in detecting multiple objects using LiDAR data obtained from a single, fixed location. Real-time experiments are conducted with human subjects navigating predefined paths. Three individuals move within an environment, while LiDAR, fixed at the center, dynamically tracks and identifies their locations at multiple instances. Results demonstrate that a single, strategically positioned LiDAR can adeptly detect objects in motion around it. Furthermore, this study provides a comparison of various regression techniques for predicting bounding box coordinates. Gaussian process regression (GPR), combined with particle swarm optimization (PSO) for prediction, achieves the lowest prediction mean square error of all the regression techniques examined at 0.01. Hyperparameter tuning of GPR using PSO significantly minimizes the regression error. Results of the experiment pave the way for its extension to various real-time applications such as crowd management in malls, surveillance systems, and various Internet of Things scenarios.

Robotic manipulation requires object pose knowledge for the objects of interest. In order to perform typical household chores, a robot needs to be able to estimate 6D poses for objects such as water glasses or salad bowls. This is especially difficult for glass objects, as for these, depth data are mostly disturbed, and in RGB images, occluded objects are still visible. Thus, in this paper, we propose to redefine the ground-truth for training RGB-based pose estimators in two ways: (a) we apply a transparency-aware multisegmentation, in which an image pixel can belong to more than one object, and (b) we use transparency-aware bounding boxes, which always enclose whole objects, even if parts of an object are formally occluded by another object. The latter approach ensures that the size and scale of an object remain more consistent across different images. We train our pose estimator, which was originally designed for opaque objects, with three different ground-truth types on the ClearPose dataset. Just by changing the training data to our transparency-aware segmentation, with no additional glass-specific feature changes in the estimator, the ADD-S AUC value increases by 4.3%. Such a multisegmentation can be created for every dataset that provides a 3D model of the object and its ground-truth pose. [ABSTRACT FROM AUTHOR]

The environmental perception system is the foundation of unmanned driving systems and also the fundamental guarantee of the safety and intelligence of unmanned vehicles. The obstacle hazard identification technology is the core of the environment perception system, and it is also the basic condition for the autonomous driving of unmanned vehicles. In view of the complexity of obstacle danger identification, this research paper designs an improved Density-Based Spatial Clustering of Applications with Noise (DBSCAN) Algorithm for hazard recognition of obstacles in unmanned scenes through a systematic approach. First, it highlights the significance of morphological component analysis in identifying non-smooth regions within images where obstacles are likely to be present. Second, it introduces a novel approach for core point definition by identifying an optimal MinDensity value based on the curvature of the density distribution curve. Third, it addresses variations in density sequences through smoothing and normalization. Finally, it constructs an improved DBSCAN Algorithm for hazard recognition of obstacles in unmanned scenes. It addresses limitations in the traditional DBSCAN by refining the core point definition using an adaptive density threshold. It identifies the "elbow point" in density distribution, enhancing its ability to distinguish density states. Additionally, it incorporates density curve smoothing, normalization, and a merger step for handling stationary objects. The results show that it has high accuracy (95.6%), precision (96.3%), recall (94.5%), and F-Score (95.4%), as well as increased consistency (92.5%) and dependability (93.2%). It also has fast real-time data processing, lasting only 0.12 s, making it an excellent choice for obstacle detection and unmanned hazard avoidance. [ABSTRACT FROM AUTHOR]

The construction of bounding boxes is an effective method to reduce the complexity of collision detection in virtual assembly. In response to the shortcomings of creating AABB bounding boxes by default for Box components in the X3D environment, a method for creating OBB bounding boxes in this environment is studied. By calculating Euler angles and combining with Transform components, the construction of OBB bounding boxes in this environment is implemented. In order to improve the efficiency of constructing bounding boxes in X3D environment, a hybrid adaptive bounding box generation algorithm based on AABB and OBB is proposed, which adds judgment on the degree of inclination of objects and enables it to adaptively select AABB or OBB methods to construct bounding boxes based on the geometric features of the 3D model. The research results show that when the included angle threshold is set to 15°, this algorithm reduces the total generation time of bounding box for the entire teapot by 5.61% compared to the pure OBB method, and the total volume by 1.53% compared to the pure AABB method. This algorithm combines the fast generation speed of AABB bounding boxes and the good tightness of OBB bounding boxes, making it an effective bounding box construction algorithm. [ABSTRACT FROM AUTHOR]

In view of the problem that when lidar collects data, due to the negative impact of external interference factors and scanning accuracy, the spatial density of the collected point cloud data will vary greatly, and there will be a lot of noise and holes, so that the analysis results can not directly describe the model of the actual object, a point cloud data processing algorithm was designed based on binary grid occupation. First, the point cloud after segmentation was clustered by binary grid to reduce the dimension, and then the point cloud was mapped to the grid cell to achieve rapid aggregation of different object point clouds. Finally, the point cloud was rotated based on the main direction found Establish a closely followed obstacle bounding box. The experimental results show that this method can improve the operation speed while ensuring the clustering accuracy. Its bounding box can accurately reflect the size of obstacles, and has good real-time and follow-up performance. It provides reliable information for the mobile robot to avoid obstacles autonomously. [ABSTRACT FROM AUTHOR]

A well-known problem of visual object tracking is the difficulty of accurately estimating the object trajectory under conditions of environmental disturbances in the bounding box (BB) of a video camera. In this paper, we consider BB variations as Gaussian-Markov colored measurement noise (CMN). In order to perform accurate tracking in the presence of CMN, we use measurement differencing and develop a robust general unbiased finite impulse response (GUFIR) filter and use the general Kalman filter (GKF) as a benchmark. The GUFIR and GKF algorithms are tested by the “Car4” benchmark. It is shown that, in terms of the tracking precision and under the heavy disturbance with the $0.65 \leqslant \Psi \leqslant 0.95$ coloredness factor, the best tracking performance is achieved using the robust GUFIR filter. When $\Psi < 0.6$ , the GUFIR and GKF algorithms perform near equally. In the extreme point of $\Psi = 1.0$ , where the Gauss-Markov CMN loses the stationarity, both algorithms provide zero precision and become inefficient. In general, it is concluded that the GUFIR filter, which ignores any zero mean disturbance and initial values, is much more suitable for applications in visual object tracking than Kalman-like algorithms relying on complete object information.

The use of Deep learning techniques in the field of Marine Science has become popular in recent years. For instance, many works propose the application of instance segmentation neural networks (in particular, Mask R-CNN) for detection and classification of fish in underwater images. The performance of these learning-based approaches depends heavily on the volume of data used for training, which, in the case of instance segmentation models for fish detection, implies that human experts must label and mark the shapes of all the fish appearing in a vast amount of underwater images. This is an enormously time-consuming task that we seek to alleviate in this paper. We propose a training strategy that combines manual and semi-automatic annotations. The latter are obtained in a weakly-supervised manner: the bounding box that contains the fish is manually selected, but its shape is automatically obtained thanks to a pretrained encoder-decoder segmentation network. Several popular architectures for this encoder-decoder network are examined. This strategy permits to reduce drastically the annotation cost for instance segmentation, at the expense of a small drop in performance with respect to the use of fully manual annotations. We show that a balance can be achieved between the segmentation performance and the time used to collect the training data by using the proposed strategy.

In Senegal, stock recovery and fish classification are based on manual data collection, and the fish caught by the fishery are not often declared. What's more, data collection suffers from a lack of tools for monitoring and counting fish caught at fishing docks. Researchers have carried out studies on the fishery in Senegal, but data collection is almost non-existent. Moreover, there is no local fisheries database or automatic detection and counting algorithm. In this paper, a semantic segmentation algorithm is proposed using intelligent systems for the collection of fishery catches, for the formation of the local database. The data are collected by taking images of fish at the Soumbédioune fishing wharf in Senegal, and are completed with the Fishbase database. They were applied to the algorithm and resulted in a segmented dataset with masks. This constitutes our local database. The database is used with YOLO v8. The latter is very important for detecting images with bounding boxes in order to train the model. The results obtained are very promising for the proposed automatic poison detection and counting model. For example, the recall-confidence scores translate into bounding box performance with scores ranging from 0.01 to 0.75, confirming the performance of this model with bounding boxes. [ABSTRACT FROM AUTHOR]

Leukemia can be detected by an abnormal rise in the number of immature lymphocytes and by a decrease in the number of other blood cells. To diagnose leukemia, image processing techniques are utilized to examine microscopic peripheral blood smear (PBS) images automatically and swiftly. To the best of our knowledge, the initial step in subsequent processing is a robust segmentation technique for identifying leukocytes from their surroundings. The paper presents the segmentation of leukocytes in which three color spaces are considered in this study for image enhancement. The proposed algorithm uses a marker-based watershed algorithm and peak local maxima. The algorithm was used on three different datasets with various color tones, image resolutions, and magnifications. The average precision for all three-color spaces was the same, i.e. 94% but the Structural Similarity Index Metric (SSIM) and recall of HSV were better than other two. The results of this study will aid experts in narrowing down their options for segmenting leukemia. Based on the comparison, it was concluded that when the colour space correction technique is used, the accuracy of the proposed methodology improves. [ABSTRACT FROM AUTHOR]

Background. Figures and captions in medical documentation contain important information. As a result, researchers are becoming more interested in obtaining published medical figures from medical papers and utilizing the captions as a knowledge source. Methods. This work introduces a unique and successful six-fold methodology for extracting figure-caption pairs. The A-torus wavelet transform is used to retrieve the first edge from the scanned page. Then, using the maximally stable extremal regions connected component feature, text and graphical contents are isolated from the edge document, and multi-layer perceptron is used to successfully detect and retrieve figures and captions from medical records. The figure-caption pair is then extracted using the bounding box approach. The files that contain the figures and captions are saved separately and supplied to the end useras theoutput of any investigation. The proposed approach is evaluated using a self-created database based on the pages collected from five open access books: Sergey Makarov, Gregory Noetscher and Aapo Nummenmaa's book "Brain and Human Body Modelling 2021", "Healthcare and Disease Burden in Africa" by Ilha Niohuru, "All-Optical Methods to Study Neuronal Function" by Eirini Papagiakoumou, "RNA, the Epicenter of Genetic Information" by John Mattick and Paulo Amaral and "Illustrated Manual of Pediatric Dermatology" by Susan Bayliss Mallory, Alanna Bree and Peggy Chern. Results. Experiments and findings comparing the new method to earlier systems reveal a significant increase in efficiency, demonstrating the suggested technique's robustness and efficiency. [ABSTRACT FROM AUTHOR]

The technological development in today's era has unlocked the measures to propose new applications for the fruit industry. Automation boosts the economic growth and productivity of the country. Fruit quality detection in complex backgrounds using an automated system is significant for this sector. Fruit sorting has an impact on the export market and quality evaluation. One of the crucial qualities of grading fruits is their appearance, which affects their market value and the choice of the consumers. The manual sorting and inspection method takes a long time and is more tedious and exhaustive. Hence, an automated system is required to evaluate fruit, detect defects, and sort them based on their quality. Deep learning algorithms have highly influenced the area of object detection. Mask R-CNN and YOLOv5 are two object detection algorithms that have been experimented. YOLOv5 outperforms the Mask R-CNN approach when real-time object detection is required. The fruit identification and quality detection model is developed based on the YOLOv5 object detection system in the proposed work. The dataset includes 10,545 images of four different fruits, i.e., apple, banana, orange, and tomato, based on their quality. The model works in two phases. In phase 1, fruit is identified, and in phase 2, fruit quality detection is performed. The mosaic augmentation on the dataset has been applied for phase 1 training resulting in high detection performance and a robust system. The model classifies the fruit, and then the predicted image is passed to phase 2 for corresponding fruit quality detection. The mAP value of phase 1 is 92.80%. For phase 2, the mAP values for apple and banana quality detection models are 99.60% and 93.1%, respectively. The mAP values are 96.70% and 95% for orange and tomato quality detection models. The results show that the proposed method could realize fruit identification and quality detection on the validation dataset. The samples have been passed to show the real-time performance of the system. The efficiency of the trained model has been validated in different scenarios, including simple, complex, low-quality camera inputs. The fruit identification and quality detection model has been compared with several state-of-the-art detection methods, and the results are very encouraging. [ABSTRACT FROM AUTHOR]

To improve the speed and dynamic adaptability of robotic arm trajectory planning, a collision-free optimal trajectory planning method combining non-uniform adaptive time meshing and bounding box collision detection was proposed. First, the dynamics and objective function of the asymmetric industrial robotic arm with three degrees of freedom (DOF) was formulated in the form of the dynamic optimization problem. Second, the control vector parameterization (CVP) was improved to enhance the computational performance of the problem. The discrete grid was adaptively adjusted according the trend of control variables. Then, a quick and effective collision detection strategy was used to avoid obstacles and to speed up calculation efficiency. The non-collision constraint is built by transforming the collision detection into the distance between two points, and then is combined into the dynamic optimization problem. The solution of the new optimization problem with the improved CVP leads to the higher calculation performance and the avoidance of obstacles. Lastly, the Siemens Manutec R3 robotic arm is taken as an example to verify the effectiveness of the planning method. The approach not only reduces computation time but also maintains accurate calculations, so that optimal trajectory can be selected from symmetric paths near the obstacles. When weights were set as λ1 = λ2 = 0.5, the solution efficiency was improved by 33%, and the minimum distance between the robotic arm and obstacle could be 0.08 m, which ensured that there was no collision. [ABSTRACT FROM AUTHOR]