Multi-target Tracking Algorithm Research Articles

A Multi-Active and Multi-Passive Sensor Fusion Algorithm for Multi-Target Tracking in Dense Group Clutter Environments

Multi-target tracking (MTT) of multi-active and multi-passive sensor (MAMPS) systems in dense group clutter environments is facing significant challenges in measurement fusion. Due to the difference in measurement information characteristics in MAMPS fusion, it is difficult to effectively correlate and fuse different types of sensors’ measurements, leading to difficulty in taking full advantage of various types of sensors to improve target tracking accuracy. To this end, we present a novel MAMPS fusion algorithm, which is based on centralized measurement association fusion (MAF) and distributed deep neural network (DNN) track fusion, named the MAMPS-MAF-DNN algorithm. Firstly, to reduce the impact of the dense group clutter, a clutter pre-processing algorithm is elaborated, which combines the advantages of the CFDP (cluster by finding density peaks) and double threshold screening algorithms. Then, for the single-active and multi-passive sensor (SAMPS) system, a centralized MAF algorithm based on angle information is developed, called the SAMPS-MAF algorithm. Finally, the SAMPS-MAF algorithm is extended to the MAMPS system within the DNN framework, and the complete MAMPS-MAF-DNN algorithm is proposed. Experimental results indicate that, compared to the existing MAF and covariance intersection (CI) fusion algorithms, the proposed MAMPS-MAF-DNN algorithm can fully combine the advantages of multi-active and multi-passive sensors, efficiently reduce the computational complexity, and obviously improve the tracking accuracy.

Multi-Target Vehicle Tracking Algorithm Based on Improved DeepSORT.

In this paper, we address the issues of insufficient accuracy and frequent identity switching in the multi-target tracking algorithm DeepSORT by proposing two improvement strategies. First, we optimize the appearance feature extraction process by training a lightweight appearance extraction network (OSNet) on a vehicle re-identification dataset. This makes the appearance features better suited for the vehicle tracking model required in our paper. Second, we improve the metric of motion features by using the original IOU distance metric or GIOU metrics. The optimized tracking algorithm using GIOU achieves effective improvements in tracking precision and accuracy. The experimental results show that the improved vehicle tracking models MOTA and IDF1 are enhanced by 4.6% and 5.9%, respectively. This allows for the stable tracking of vehicles and reduces the occurrence of identity switching phenomenon to a certain extent.

An algorithm for multi-target tracking in low-signal-to-clutter-ratio underwater acoustic scenes

To simplify the computational complexity of the Joint Probabilistic Data Association (JPDA) algorithm for multi-target tracking in low-signal-to-clutter-ratio underwater acoustic scenes, and to solve the problem of unstable tracking caused by poor measurement accuracy of active sonar, a trajectory information-based JPDA optimization algorithm (TJPDA) is designed to meet the real-time and stability requirements of practical engineering. The false alarm rate is reduced by accumulating correlated target trajectories in multiple frames, and the influence factor is introduced to modify the correlation probability and simplify the confirmation matrix splitting process. This effectively solves the problem of information combination explosion and trajectory divergence caused by adjacent target interference in low signal-to-clutter-ratio underwater acoustic scenes. The simulation results show that compared with the JPDA algorithm, the TJPDA algorithm not only improves target-tracking accuracy but also reduces computation time, making it more suitable for practical engineering applications.

Detection and Tracking of Low-Frame-Rate Water Surface Dynamic Multi-Target Based on the YOLOv7-DeepSORT Fusion Algorithm

This study aims to address the problem in tracking technology in which targeted cruising ships or submarines sailing near the water surface are tracked at low frame rates or with some frames missing in the video image, so that the tracked targets have a large gap between frames, leading to a decrease in tracking accuracy and inefficiency. Thus, in this study, we proposed a water surface dynamic multi-target tracking algorithm based on the fusion of YOLOv7 and DeepSORT. The algorithm first introduces the super-resolution reconstruction network. The network can eliminate the interference of clouds and waves in images to improve the quality of tracking target images and clarify the target characteristics in the image. Then, the shuffle attention module is introduced into YOLOv7 to enhance the feature extraction ability of the target features in the recognition network. Finally, Euclidean distance matching is introduced into the cascade matching of the DeepSORT algorithm to replace the distance matching of IOU to improve the target tracking accuracy. Simulation results showed that the algorithm proposed in this study has a good tracking effect, with an improvement of 9.4% in the improved YOLOv7 model relative to the mAP50-95 value and an improvement of 13.1% in the tracking accuracy in the DeepSORT tracking network compared with the SORT tracking accuracy.

Rapeseed Flower Counting Method Based on GhP2-YOLO and StrongSORT Algorithm.

Accurately quantifying flora and their respective anatomical structures within natural ecosystems is paramount for both botanical breeders and agricultural cultivators. For breeders, precise plant enumeration during the flowering phase is instrumental in discriminating genotypes exhibiting heightened flowering frequencies, while for growers, such data inform potential crop rotation strategies. Moreover, the quantification of specific plant components, such as flowers, can offer prognostic insights into the potential yield variances among different genotypes, thereby facilitating informed decisions pertaining to production levels. The overarching aim of the present investigation is to explore the capabilities of a neural network termed GhP2-YOLO, predicated on advanced deep learning techniques and multi-target tracking algorithms, specifically tailored for the enumeration of rapeseed flower buds and blossoms from recorded video frames. Building upon the foundation of the renowned object detection model YOLO v8, this network integrates a specialized P2 detection head and the Ghost module to augment the model's capacity for detecting diminutive targets with lower resolutions. This modification not only renders the model more adept at target identification but also renders it more lightweight and less computationally intensive. The optimal iteration of GhP2-YOLOm demonstrated exceptional accuracy in quantifying rapeseed flower samples, showcasing an impressive mean average precision at 50% intersection over union metric surpassing 95%. Leveraging the virtues of StrongSORT, the subsequent tracking of rapeseed flower buds and blossom patterns within the video dataset was adeptly realized. By selecting 20 video segments for comparative analysis between manual and automated counts of rapeseed flowers, buds, and the overall target count, a robust correlation was evidenced, with R-squared coefficients measuring 0.9719, 0.986, and 0.9753, respectively. Conclusively, a user-friendly "Rapeseed flower detection" system was developed utilizing a GUI and PyQt5 interface, facilitating the visualization of rapeseed flowers and buds. This system holds promising utility in field surveillance apparatus, enabling agriculturalists to monitor the developmental progress of rapeseed flowers in real time. This innovative study introduces automated tracking and tallying methodologies within video footage, positioning deep convolutional neural networks and multi-target tracking protocols as invaluable assets in the realms of botanical research and agricultural administration.

Improved Particle Filter Algorithm for Multi-Target Detection and Tracking.

In modern radar detection systems, the particle filter technique has become one of the core algorithms for real-time target detection and tracking due to its good nonlinear and non-Gaussian system state estimation capability. However, when dealing with complex dynamic scenes, the traditional particle filter algorithm exposes obvious deficiencies. The main expression is that the sample degradation is serious, which leads to a decrease in estimation accuracy. In multi-target states, the algorithm is difficult to effectively distinguish and stably track each target, which increases the difficulty of state estimation. These problems limit the application potential of particle filter technology in multi-target complex environments, and there is an urgent need to develop a more advanced algorithmic framework to enhance its robustness and accuracy in complex scenes. Therefore, this paper proposes an improved particle filter algorithm for multi-target detection and tracking. Firstly, the particles are divided into tracking particles and searching particles. The tracking particles are used to maintain and update the trajectory information of the target, and the searching particles are used to identify and screen out multiple potential targets in the environment, to sufficiently improve the diversity of the particles. Secondly, the density-based spatial clustering of applications with noise is integrated into the resampling phase to improve the efficiency and accuracy of particle replication, so that the algorithm can effectively track multiple targets. Experimental result shows that the proposed algorithm can effectively improve the detection probability, and it has a lower root mean square error (RMSE) and a stronger adaptability to multi-target situation.

PMBM-based multi-target tracking under measurement merging

Most multi-target tracking (MTT) algorithms have been developed based on the assumption that a measurement can be produced by at most one target. In contrast, the situation that multiple targets produce a single measurement has been addressed relatively rarely. Such a problem arises in the application where targets are close to each other in the measurement space and thus occupy the same resolution bin. To handle such a problem, this paper proposes a Poisson multi-Bernoulli mixture (PMBM) based approach to perform MTT under measurement merging. The proposed method shares the same prediction step as the standard PMBM filter. However, the PMBM density is not a conjugate prior to the multi-target likelihood with merged measurements, and an approximate PMBM density to the posterior is obtained. The Gaussian mixture based implementation strategy is introduced in this paper, and the Gibbs sampling strategy is further exploited to implement the proposed method efficiently. The performance of proposed method is verified via simulations.

Real-time tracking method for motion spatter in high-power laser welding of stainless steel plate based on a lightweight deep learning model

The high-power laser welding of stainless steel is prone to spatter defects. The dynamic behavior of motion spatter is closely related to the welding state, but the relevant features are difficult to obtain. This paper extensively describes a motion spatter tracking method based on a lightweight deep learning model and a multi-target tracking algorithm. Firstly, an in-situ observation platform with a high-speed camera-based high spatiotemporal resolution for motion spatters is established. This platform can record the generation and motion process of different spatters during welding. To reduce the interferences and highlight the spatter characteristics, a linear point operation-based image low-brightness metal vapor plume removal method is also proposed. Then, a deep learning-based lightweight spatter detection model is established, and an automatic spatter label generation method is proposed to quickly obtain a sufficient amount of highly discriminative training data to train the established model comprehensively. Compared with the basic model, the validation results show that the lightweight spatter detection model runs 5.11 times faster, with a spatter detection accuracy of 96.71%. Finally, a motion spatter tracking method based on the DeepSORT algorithm is proposed to distinguish different spatters and associate the same spatters. The spatter size, velocity, quantity, and other features are also able to be accurately extracted via this method, achieving spatter defect monitoring. In summary, this study describes a novel and robust method for motion spatter real-time tracking during laser welding and this method can provide reliable data support for accurate and fast process condition monitoring as well as weld quality assessment.

Design and Simulation of Multi Unmanned Boat Cooperative Obstacle Avoidance System Based on 5G Edge Computing

Background:: Compared to single unmanned boat, multi unmanned boats have more flexible mobility and efficient task completion capabilities, which can effectively expand the types of tasks. However, the traditional independent path planning and obstacle avoidance methods of unmanned boats are difficult to meet the requirements of collaborative operation among multiple unmanned boats due to the lack of information exchange. Objective:: According to the actual demand of multi unmanned boats' cooperative operation, a method of multi unmanned boats cooperative obstacle avoidance based on 5G edge computing is proposed to realize the unified planning and scheduling of multi unmanned boats. Methods:: Firstly, 5G technology and Kubeedge edge computing tools are used to build a multi unmanned boat collaborative obstacle avoidance system based on cloud, edge and end collaboration, and the Kubeedge edge computing platform was optimized by optimizing communication strategies, building a highly available Kubeedge cluster, building a Harbor image center, and using Web management interfaces further to improve the reliability and stability of the system. Secondly, the YOLOR-Deepport multi-target recognition and tracking algorithm based on cloud, edge and end collaborative network is used to complete the recognition and tracking tasks of obstacle targets, and a set of EECBS path planning methods based on the Kubedge centralized control platform is designed to plan collision-free and efficient paths for each unmanned boat in realtime. Finally, the effectiveness of the system was verified through simulation experiments. Results:: The experimental results show that compared to the traditional autonomous planning obstacle avoidance method for unmanned boats, the collaborative planning obstacle avoidance method proposed in this paper can exhibit excellent performance in dense and narrow scenarios, with a more reasonable navigation path, a range reduction of 20% - 50%, and higher safety. Conclusion:: The results show that the cooperative obstacle avoidance system based on 5G edge computing designed in the paper is feasible, and it can effectively realize the cooperative path planning and obstacle avoidance of multi unmanned boats.

The multiple pairwise Markov chain model-based labeled multi-Bernoulli filter

Most of existing multi-target tracking (MTT) algorithms, which are rooted in random finite set theory, generally rely on two hypotheses, i.e., the single dynamic model hypothesis and the hidden Markov chain (HMC) hypothesis, and the HMC hypothesis requires the target state to conform to a Markov process and the detection process to be independent. Unfortunately, these hypotheses may not always hold at the same time in many practical situations. Therefore, it is important to study the MTT algorithms in such scenarios when the HMC hypothesis and the single dynamic model hypothesis fail simultaneously. As a result, this paper presents a multiple model MTT algorithm, which is designed to tackle the MTT problem effectively in scenarios where both hypotheses are invalid. Firstly, when the HMC hypothesis is not satisfied, an MTT algorithm was presented based on pairwise Markov chain (PMC) and the labeled multi-Bernoulli filter (PMC-LMB). Secondly, in case that both hypotheses are not met, a multiple model MTT algorithm was proposed by extending the previously presented PMC-LMB filter to multiple PMC model case. Finally, extensive simulation was done to demonstrate the efficiency of the presented algorithms.

Research on Multi-target Tracking Algorithm Based on Track Segment Association

In the process of multi-target tracking, the complex underwater acoustic channel, target model mismatch, target strong maneuvering and other factors are easy to cause track interruption and rerun, which have a bad influence on the situation assessment and tactical decision of the Unmanned sonar system. Aiming at the problem of track interruption, this paper integrates the track adhesion algorithm into the active multi-target tracking algorithm. Simulation results show that the track adhesion algorithm can effectively improve the correct track association rate and track lifetime. By comparing the effect of the new and old methods with the sea test data, the track adhesion can effectively solve the problem of track interruption in the case of target track maneuvering and improve the track life.

Intelligent Traffic System: Yolov8 and Deepsort in Car Detection, Tracking and Counting

Abstract: As the number of vehicles on the road increases, it becomes essential to quickly and accurately identify them to better detect traffic congestion and improve traffic management. DeepSORT (simple online and real-time tracking with a deep association metric) multitarget tracking algorithm in vehicle tracking. Due to the strong dependence of the DeepSORT algorithm on target detection, a YOLOv8 vehicle detection algorithm was proposed based on YOLOv7, which provides accurate and fast vehicle detection data to the DeepSORT algorithm. DeepSORT introduces deep learning into the SORT algorithm by adding an appearance descriptor to reduce identity switches, making tracking more efficient. DeepSORT uses a better association metric that combines both motion and appearance descriptors. DeepSORT can be defined as the tracking algorithm that tracks objects not only based on the velocity and motion of the object but also the appearance of the object. This vehicle detection also uses the DeepSORT algorithm to help count the number of vehicles that pass in the video effectively. YOLOv8 and the DeepSORT algorithm collaborate to identify and follow vehicles creating a model, for vehicle recognition that showcases their ability to track cars efficiently. From this paper, the model Yolov8 has achieved state-of-the-art results

An anti-occlusion optimization algorithm for multiple pedestrian tracking.

Frequent occlusion of tracking targets leads to poor performance of tracking algorithms. A common practice in multi-target tracking algorithms is to re-identify the occluded tracking targets, which increases the number of identity switching occurrences. This paper focuses on online multi-object tracking and designs an anti-occlusion, robust association strategy, and feature extraction model. Specifically, the least squares algorithm and the Kalman filter are used to predict the trajectory of the tracking target, while the two-way self-attention mechanism is employed to extract the features of the tracking target, as well as positive and negative samples. After the tracking target is occluded, the association strategy is used to assign the identity information from before the occlusion. The experimental results demonstrate that the algorithm proposed in this paper has achieved excellent tracking performance on the MOT dataset.

Deep learning and multi-modal fusion for real-time multi-object tracking: Algorithms, challenges, datasets, and comparative study

Real-time multi-object tracking (MOT) is a complex task involving detecting and tracking multiple objects. After the objects are detected, they are assigned markers, and their trajectories are tracked in real-time. The scientific community is intrigued by the possibilities of utilizing MOT technology in the context of smart cities. Their primary focus lies in the domains of intelligent transportation, detection of vehicles and pedestrians, crowd surveillance, and public safety. Deep learning techniques have been developed in recent years to effectively tackle the challenges of real-time MOT tasks and enhance tracking performance. Environmental perception within smart traffic applications relies heavily on sensor data fusion. In traffic scenarios, a thoughtful approach involves utilizing a combination of sensors and cameras to detect and track targets while gathering valuable data effectively. However, it faces challenges when it comes to detecting and tracking objects that are in motion, have complex changes in appearance, or are in crowded scenes. This paper explores the foundational standard for real-time Multiple Object Tracking tasks. We prioritize the examination of quantitative measures by conducting a comprehensive analysis of widely utilized benchmark datasets and metrics. This study also investigates established embedding techniques and multi-modal fusion methods within real-time multi-target tracking algorithms. Each strategy will be classified and assessed according to a predefined set of principles. The paper presents a comprehensive analysis and visual representation of various MOT strategies. Finally, this paper aims to present an overview of the current challenges faced by the MOT mission, as well as the potential objectives that lie ahead.

Application analysis of multi-target tracking algorithm based on BSIC-re ID and spatio-temporal constraints in basketball games

Application analysis of multi-target tracking algorithm based on BSIC-re ID and spatio-temporal constraints in basketball games

Robust cardinalized probability hypothesis density filter based underwater multi-target direction-of-arrival tracking with uncertain measurement noise

Direction-of-arrival (DOA) tracking of underwater targets is an important research topic in sonar signal processing. Considering the underwater DOA tracking is a typical multi-target scenario with uncertain measurement noise caused by unknown underwater environment, a robust underwater multi-target DOA tracking method with uncertain measurement noise is proposed by implementing a spatial matrix filtering (SMF) technique in the framework of the cardinalized probability hypothesis density (CPHD) filter. Firstly, the kinematic model of underwater targets and the measurement model based on the received signal of the hydrophone array (array signal) are established. Thus, the multi-target DOA tracking algorithm is derived by using the CPHD filter based on the established model. Moreover, to avoid the effect of the uncertain measurement noise, the SMF technique is implemented to process the array signal to extract the signal in the bearing angle of the target. Then, the output of the SMF is substituted into the CPHD filter. In this manner, the robust underwater multi-target DOA tracking method for uncertain measurement noise scenario is proposed. Finally, comprehensive simulations and experimental data processing are performed to prove the robustness of the proposed method in the scenario of multi-target DOA tracking with uncertain measurement noise. Compared with the existing DOA estimation and DOA tracking methods, the proposed methods significantly improve the robustness and the precision of the DOA of the underwater target.

Improved UAV-to-Ground Multi-Target Tracking Algorithm Based on StrongSORT.

Unmanned aerial vehicles (UAV) are essential for aerial reconnaissance and monitoring. One of the greatest challenges facing UAVs is vision-based multi-target tracking. Multi-target tracking algorithms that depend on visual data are utilized in a variety of fields. In this study, we present a comprehensive framework for real-time tracking of ground robots in forest and grassland environments. This framework utilizes the YOLOv5n detection algorithm and a multi-target tracking algorithm for monitoring ground robot activities in real-time video streams. We optimized both detection and re-identification networks to enhance real-time target detection. The StrongSORT tracking algorithm was selected carefully to alleviate the loss of tracked objects due to factors like camera jitter, intersecting and overlapping targets, and smaller target sizes. The YOLOv5n algorithm was used to train the dataset, and the StrongSORT tracking algorithm incorporated the best-trained model weights. The algorithm's performance has greatly improved, as demonstrated by experimental results. The number of ID switches (IDSW) has decreased by sixfold, IDF1 has increased by 7.93%, and false positives (FP) have decreased by 30.28%. Additionally, the tracking speed has reached 38 frames per second. These findings validate our algorithm's ability to fulfill real-time tracking requisites on UAV platforms, delivering dependable resolutions for dynamic multi-target tracking on land.

Multi-target tracking algorithm in aquaculture monitoring based on deep learning

In order to analyze fish behavior, real-time underwater fish monitoring is essential. This work presents an underwater multi-target tracking algorithm for aquaculture monitoring, utilizing deep learning techniques. Underwater images are acquired using an underwater robot, and the images are defogged using the multi-scale Retinex algorithm based on HSV space. To enhance object detection performance, GhostNetv2 was integrated into the You Only Look Once version 5 (YOLOv5) object detection algorithm, supplemented by the addition of the Coordinate Attention (CA) module, resulting in the development of GN-YOLOv5. For the tracking algorithm, the more accurate Generalized Intersection over Union (GIoU) method was incorporated into the StrongSORT tracking algorithm. Moreover, to achieve more precise target tracking, a fish re-identification model was established. The proposed algorithms were evaluated through experiments conducted on various datasets, including VOC 2012, MOT16, and self-built datasets. The results demonstrate notable improvements: the GN-YOLOv5 model showed a 32.91% reduction in parameters and a 3.22% increase in precision. Furthermore, the enhanced StrongSORT algorithm exhibited a 3.84% increase in MOTA, a 28.00% reduction in IDS, and a speed boost of 7 FPS, leading to stable and accurate multi-target fish tracking.

Multi target tracking algorithm based on anchor free detection

In order to alleviate the problems of frequent identity label switching caused by mutual occlusion between targets and missed detection caused by small target sizes in multi-target tracking, this paper proposes a multi-target tracking algorithm based on anchor-free detection. This algorithm first uses YOLOX for detection and embeds the ReID algorithm to enhance pedestrian recognition. Then, it uses the DeepSORT algorithm for tracking and correcting abnormal trajectory algorithms for occluded targets. Finally, experiments are conducted on both the MOT16 dataset and the self-made dataset. This algorithm has higher tracking accuracy than other algorithms, and its tracking performance for small targets has also been effectively improved.

Research on real-time perception method of key targets in the comprehensive excavation working face of coal mine

Aiming at the real-time perception problem of key target objects caused by harsh environmental factors of high dust, low illumination, motion blur, and multi-target mixing in the comprehensive excavation working face of coal mine, a multi-target detection and tracking algorithm based on DDEB-YOLOv5s and StrongSORT is proposed. First, the YOLOv5s model is improved by using C3-Dense module, decoupled head, ECIoU loss function, and weighted bi-directional feature pyramid network to enhance the detection performance of the model in complex backgrounds of coal mine and complete the design of the DDEB-YOLOv5s multi-target detection network. Second, the DDEB-YOLOv5s algorithm is used as a detector and combined with the StrongSORT tracking algorithm to track critical equipment and miners in the complex context of coal mine. Experiments were conducted on the dataset of comprehensive excavation working face, and the experimental results show that the proposed DDEB-YOLOv5s has the best integrated detection performance compared with other YOLO series target detection algorithms, and its mean value of detection accuracy reaches 91.7%, which is 4.9% higher than that of the original YOLOv5s model. In addition, compared to the three tracking models, (YOLOv7-tiny)-(BoT-SORT), YOLOv5s-DeepSORT, and YOLOv8s-Bytetrack, the (DDEB-YOLOv5s)-StrongSORT model has the best tracking performance (with a mean tracking accuracy of 94.2%) and the least number of identifier switches. Finally, the real-time perception method proposed in this study for the key target of the coal mine working face can provide new technical support and effective guarantee for coal mine safety production.