Multiple Target Tracking Research Articles

Deep-Reinforcement-Learning-Based Radar Parameter Adaptation for Multiple-Target Tracking

Deep-Reinforcement-Learning-Based Radar Parameter Adaptation for Multiple-Target Tracking

One-Stage Anchor-Free Online Multiple Target Tracking with Deformable Local Attention and Task-Aware Prediction.

The tracking-by-detection paradigm currently dominates multiple target tracking algorithms. It usually includes three tasks: target detection, appearance feature embedding, and data association. Carrying out these three tasks successively usually leads to lower tracking efficiency. In this paper, we propose a one-stage anchor-free multiple task learning framework which carries out target detection and appearance feature embedding in parallel to substantially increase the tracking speed. This framework simultaneously predicts a target detection and produces a feature embedding for each location, by sharing a pyramid of feature maps. We propose a deformable local attention module which utilizes the correlations between features at different locations within a target to obtain more discriminative features. We further propose a task-aware prediction module which utilizes deformable convolutions to select the most suitable locations for the different tasks. At the selected locations, classification of samples into foreground or background, appearance feature embedding, and target box regression are carried out. Two effective training strategies, regression range overlapping and sample reweighting, are proposed to reduce missed detections in dense scenes. Ambiguous samples whose identities are difficult to determine are effectively dealt with to obtain more accurate feature embedding of target appearance. An appearance-enhanced non-maximum suppression is proposed to reduce over-suppression of true targets in crowded scenes. Based on the one-stage anchor-free network with the deformable local attention module and the task-aware prediction module, we implement a new online multiple target tracker. Experimental results show that our tracker achieves a very fast speed while maintaining a high tracking accuracy.

Time‐varying group formation tracking control for multi‐agent systems using distributed multi‐sensor multi‐target filtering with intermittent observations

AbstractTime‐varying group formation tracking control problems for multi‐agent systems are investigated based on distributed multi‐sensor multi‐target filtering with intermittent observations. First, in order to estimate the states of multiple targets under the phenomenon of intermittent observations accurately, a distributed multi‐sensor multi‐target filtering algorithm is proposed based on cubature Kalman filtering. Second, a time‐varying group formation tracking protocol is designed for multi‐agent systems by using the state estimations obtained from the filtering algorithm and the neighboring interaction. The protocol enables multi‐agent systems to form time‐varying subformations and track multiple targets in the same subgroups, respectively. Third, the boundedness of the error covariance matrices is proved under the condition that the observation probability is higher than the minimum threshold. Then the estimation errors of the filtering algorithm are proved to be stochastically bounded by introducing a stochastic process. Furthermore, the boundedness of the group formation tracking errors is proved. Finally, a numerical example is used to verify the performance of the proposed algorithm and protocol.

A SIAR transmitting waveform design utilizing positive and negative random carrier frequency coding technique for coupling reduction in circular array

The Synthetic Impulse and Aperture Radar (SIAR) is an advanced radar system which utilizes different orthogonal carrier frequencies. As a four-dimensional (4-D) radar, SIAR offers numerous advantages such as protection against anti-radiation missiles (ARM), detection of stealth targets, low probability of interception (LPI), and tracking of multiple targets. However, one of the key challenges facing these radars is the coupling between range and angle measurements, where errors in range measurements can impact angle measurements and vice versa. To tackle this issue, radar designers commonly develop transmitting waveforms based on carrier frequency coding to minimize the coupling. Various methods, including random frequency coding, sequential frequency coding, and positive and negative sequential frequency coding, have been proposed to reduce the coupling between range and angle measurements for two-dimensional (2-D) linear arrays. This article focuses on extending these ideas to circular arrays (3D radar) and introduces a method known as positive and negative random carrier frequency coding. We take advantage of random code and the concept of positive-negative frequency coding. Additionally, mathematical equations and simulations demonstrate that the coupling between azimuth angle and elevation angle is independent of the frequency coding type and remains consistent across all conditions. Through our research, we demonstrate that the proposed method effectively reduces coupling by approximately 40% more than other frequency coding methods. In the proposed method, we combine output of two consecutive periods by an approach similar to what happens in pulse integration in radar, so we don't require complex computations or new hardware. This method has another advantage such as pulse-to-pulse frequency agility, enhancing protection against ARM through virtual displacement of transmitters. Circular arrangements ensure simultaneous coverage of the entire space.

LPI-based resource allocation strategy for multiple targets tracking in CMIMO radar system with array division

Considering the scenario of multiple targets equipped with interceptors in the modern electronic warfare environment, an LPI-based Resource Allocation Strategy (LPIRAS) for multiple targets tracking (MTT) in collocated MIMO (CMIMO) radar is proposed in this paper, where subarray number is exploited in the low probability of intercept (LPI) optimization problem for the first time. As the problem is formulated as a nonconvex and nonlinear optimization problem, the array power, target-beam assignment, subarray number, and subarray-beam assignment are corporately managed through a methodology incorporating the two-stage search algorithm (TSA), permutation-based target swarm generation algorithm (PTSGA) and adaptive extra subarray assignment algorithm (AESAA), resulting the LPIRAS. Finally, numerical simulations show the proposed strategy is effective and real-time. In comparison to other benchmarks, the LPIRAS achieves superior LPI performance.

Intelligent toy tracking trajectory design based on mobile cloud terminal deployment and depth-first search algorithm.

The popularization of intelligent toys enriches the lives of the general public. To provide the public with a better toy experience, we propose the intelligent toy tracking method by the mobile cloud terminal deployment and depth-first search algorithm. Firstly, we construct a toy detection model via Transformer, which realizes the positioning of toys in the image through the refined region adaptive boundary representation. Then, using these detected continuous frames, we improve the toy tracking based on a depth-first search. Long-short-term memory (LSTM) constructs the continuous frame tracking structure, and the depth-first search mechanism is embedded to realize the accurate tracking of multiple targets in continuous frames. Finally, to realize the terminal marginalization of the proposed method, this chapter proposes a lightweight model deployment method based on mobile cloud terminals to realize the maintenance of the optimal machine state of intelligent toys. The experiment proves that our proposed target method can reach the world-leading level and obtain the mAP value of 0.858. Our tracking method can also perform excellently with a MOTA value of 0.916.

Adaptive Multi-Hypothesis Marginal Bayes Filter for Tracking Multiple Targets

Tracking multiple targets in the presence of unknown number of targets, missed detection, clutter, and noise is a challenging problem. To cope with this problem, a novel approach for generating the potential birth targets was developed, a mathematical model for multiple hypotheses was established, and an adaptive multi-hypothesis marginal Bayes filter is herein proposed in terms of the established mathematical model for multiple hypotheses and the novel birth approach. This filter delivers the existence probabilities of targets and their probability density functions. It uses multiple hypotheses to solve the data association problem to form the existence probabilities of targets and their probability density functions. To obviate the requirement for prior birth models, this filter uses the observations from two consecutive time steps to establish the birth models adaptively. Its tracking performance was tested by comparing it with other adaptive filters, showing that the proposed filter is robust, and it can obtain higher tracking accuracy than other filters.

Shuffled multi-channel sparse signal recovery

Mismatches between samples and their respective channel or target commonly arise in several real-world applications. For instance, whole-brain calcium imaging of freely moving organisms, multiple-target tracking or multi-person contactless vital sign monitoring may be severely affected by mismatched sample-channel assignments. To address this issue systematically, we frame it as a signal reconstruction problem where correspondences between samples and channels are lost. Assuming a sensing matrix for the signals, we show the problem’s equivalence to a highly structured unlabeled sensing problem and establish conditions for unique recovery. This is crucial since existing unlabeled sensing theory is inapplicable and results for reconstructing shuffled multi-channel signals do not yet exist. Our results extend to continuous-time sparse signals, and we derive conditions for reconstructing shuffled sparse signals. For the two-channel case, we provide a first reconstruction method, which combines sparse signal recovery with robust linear regression, outperforming existing unlabeled sensing methods in numerical experiments. Additionally, we showcase its effectiveness in a real-world application involving calcium imaging traces. Our theory marks a significant initial step in addressing this challenging signal reconstruction problem, with potential extensions to diverse signal representations encountered in real-world problems with imprecise measurement or channel assignment.

Cognitive dual coprime frequency diverse array MIMO radar network for target discrimination and main‐lobe interference mitigation

AbstractThe authors propose a novel dual coprime frequency diverse array (FDA) multiple input multiple output (DCFDA‐MIMO) radar network design, empowered by cognitive capabilities, aimed at target discrimination and mitigation of interference present in the standalone radar systems. That is, the proposed DCFDA‐MIMO design capitalises on the complementary advantages of FDAs for target discrimination and coprime arrays for enhanced resolution, resulting in superior performance. Additionally, the proposed DCFDA‐MIMO network employs a 2D multiple signal classification algorithm to achieve high‐resolution target localisation. By incorporating cognitive techniques based on the action‐perception cycle, the proposed approach demonstrates notable improvements in multiple target detection and tracking accuracy with fewer number of antenna elements as compared to existing techniques. Furthermore, it enhances individual radar beamforming performance for interference suppression and true target detection without prior information.

Collaborative trajectory planning and transmit resource scheduling for multiple target tracking in distributed radar network system with GTAR

A reasonable resource management strategy can maximize the sensing performance of radar network system via adaptively controlling the working parameters of interest. In this paper, a collaborative trajectory planning and transmit resource scheduling (CTPTRS) strategy for multiple target tracking (MTT) in distributed radar network system with ground-based transmitters and airborne receivers (GTAR) is proposed. Firstly, the CTPTRS for the distributed radar network with GTAR (DRN-GTAR) is formulated as a non-convex optimization problem. To be more precise, the transmit power with the bandwidth of each transmitter and the trajectory of each airborne front-receiver are controlled jointly and adaptively to maximize the MTT performance. Secondly, to tackle the non-convex optimization problem, an efficient two-stage solution technique integrated with the semi-definite programming (SDP), the Latin hypercube sampling-based stochastic simulation (LHS) and the improved particle swarm optimization-tabu search (IPSOTS) is proposed. Finally, simulation results demonstrate the effectiveness of the proposed CTPTRS strategy. Furthermore, in terms of the MTT performance, the proposed strategy owns superior performance compared with six other existing baseline strategies.

Real-time and Intelligent Moving Targets Tracking based on UAV Remote Sensing Video Camera and Brain-like Computing Chips

Abstract. Moving target tracking technology based on Unmanned Aerial Vehicles (UAV) is widely used in many fields such as automatic inspection and emergency response. The existing moving target tracking methods usually have the problems of large computation and low tracking efficiency. Limited by the computing power of the UAV platform, real-time tracking and analysis of multiple targets based on the video data collected by UAV platform is a difficult task. In this paper, we proposed a novel Target Specific Filtering Tracking with Memory (TSFMTrack) method designed for UAV-based real-time tracking tasks, which involves a Tracklet Filtering Module (TFM) for capturing object appearance features and a Tracklet Matching Module (TMM) for bounding box association in each frame. By experimental comparison with other State-Of-The-Art (SOTA) methods on popular MOT and UAV tracking datasets, the TSFMTrack have shown obvious advantages in accuracy, computational efficiency and reliability. Furthermore, we deployed the TSFMTrack on the brain-inspired chip Lynchip KA200, the experimental results have shown that the TSFMTrack is effective on edge computational platform and suitable for UAV real-time tracking tasks.

Enhanced multi-model multi-scan data association and tracking algorithm via convex variational inference

Tracking multiple targets with unknown measurement-to-target association and uncertain target dynamics is a significant problem that arises in various applications such as surveillance monitoring and intelligent transportation systems. In this paper, we propose an enhanced multi-model multi-scan data association algorithm to address the problem of tracking multiple maneuvering targets. First, we use a probabilistic graphical model to represent the joint distribution of the dynamic model indices, target state, and multi-scan data association variables. This formulation transforms the inference of marginal distributions into a Bethe free energy (BFE) problem. Next, to transform the BFE problem into a convex one, we demonstrate that the BFE function can be made convex through re-weighting. Additionally, we decompose the re-weighted BFE function into a block-wise sum form. We prove that under certain regularization conditions, each block of the re-weighted BFE is convex, ensuring convergence of the primal–dual coordinate ascent algorithm to the minimum of the overall re-weighted BFE. Finally, we provide a particle implementation of the proposed algorithm, accompanied by an analysis of its complexity. Simulation results indicate that the proposed algorithm exhibits favorable performance when compared to both the single-model multi-scan algorithm and the multi-model single-scan algorithm.

A multi-target detection and position tracking algorithm based on mmWave-FMCW radar data

Detecting and tracking the position of multiple targets indoors is a challenging measurement problem due to the inherent difficulty to cluster correctly the sensor data associated to a given target and to track the position of each cluster with adequate accuracy. This problem is critical especially in rooms filled with fixed or moving objects hampering target detection and whenever the paths of different targets cross one another. In this paper, a robust Multiple Targets Tracking (MTT) algorithm exploiting the clouds of points collected from a mmWave-FMCW radar is presented. The proposed solution consists of four main steps. First, the possible outliers of a raw radar data set are removed using a neural network model. Next, the cleaned-up radar data are clustered using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. Then, a Kalman Filter (KF) is used to track the position of the centroid of each cluster. Finally, a Structured Branching Multiple Hypothesis Testing (SBMHT) algorithm is applied and updated over reasonably short time intervals to decide which detected tracks are supposed to be confirmed and which ones instead should be discarded. The proposed MTT technique was validated experimentally using the data sets collected from a 60-GHz TI IWR6843 radar platform. The reported results show that the developed algorithm, if properly tuned, is faster and returns more accurate results than other MTT techniques. In particular, the percentage of detection errors is negligible and the planar positioning accuracy is within about 30 cm with 90% probability when up to five targets move freely within the same room.

Integrated Trajectory Planning and Resource Scheduling for Multiple Target Tracking in Airborne Radar Network

Integrated Trajectory Planning and Resource Scheduling for Multiple Target Tracking in Airborne Radar Network

Joint waveform design and resource allocation strategy for cognitive radar target situation awareness

AbstractTraditional radar systems use fixed patterns and constant electromagnetic wave transmission to illuminate targets, but they often do not effectively use prior information about targets and consume significant radar resources. Cognitive radar has emerged as a way to improve resource efficiency and address these shortcomings. A joint waveform design and resource allocation strategy for cognitive radar that incorporates target situational awareness is proposed. This method integrates the interacting multiple model algorithm and the Unscented Kalman Particle Filter to achieve target situation awareness as prior knowledge. By combining the target attitude and the frequency response function of the target radar cross section at different time points in the prior knowledge, a joint beam control and power allocation strategy is formulated and transformed into an optimization problem. In addition, a cognitive pulse‐to‐pulse frequency agile waveform design method is proposed to support multiple target tracking under complex motion models. Simulation experiments demonstrate the effectiveness of this approach in obtaining accurate target situation information, achieving beam control, and optimizing power allocation. The designed waveforms can enhance radar target detection performance and improve low probability of intercept characteristics by adjusting the pulse repetition interval. This method has significant technical value.

Tiered Digital Twin-Assisted Cooperative Multiple Targets Tracking

Tiered Digital Twin-Assisted Cooperative Multiple Targets Tracking

Multidimensional morphological analysis of live sperm based on multiple-target tracking

Manual semen evaluation methods are subjective and time-consuming. In this study, a deep learning algorithmic framework was designed to enable non-invasive multidimensional morphological analysis of live sperm in motion, improve current clinical sperm morphology testing methods, and significantly contribute to the advancement of assisted reproductive technologies. We improved the FairMOT tracking algorithm by incorporating the distance and angle of the same sperm head movement in adjacent frames, as well as the head target detection frame IOU value, into the cost function of the Hungarian matching algorithm. For sperm morphology, we used the BlendMask segmentation method to segment individual sperm. SegNet was used to separate the head, midpiece, and principal piece comments from each sperm. Experienced in vivo sperm physicians confirmed a morphological accuracy percentage of 90.82%. A total of 1272 samples were collected from multiple tertiary hospitals for validation of the system, which were also evaluated by physicians. The results of our system were highly consistent with those of manual microscopy. This study realized the automated detection of progressive motility and morphology of sperm simultaneously, which is crucial for selection of morphologically normal and motile sperm for intracytoplasmic sperm injection.

Distributed Joint Detection, Tracking, and Classification via Labeled Multi-Bernoulli Filtering.

In this article, we propose a novel approach to distributed joint detection, tracking, and classification (D-JDTC) of multiple targets by means of a multisensor network. The proposed approach relies on labeled multi-Bernoulli (LMB) random finite set modeling of the multisensor state, and consists of two main tasks, that is, local filtering in each individual node and data fusion among multiple nodes. For local filtering, the LMB filter is extended to JDTC by augmenting the target state to incorporate class and mode information. Further, the well-known generalized covariance intersection and recently developed minimum information loss fusion paradigms are exploited for data fusion among sensors. The effectiveness of the resulting algorithm, called D-JDTC-LMB, is assessed via simulation experiments.

Thermal Image Tracking for Search and Rescue Missions with a Drone

Infrared thermal imaging is useful for human body recognition for search and rescue (SAR) missions. This paper discusses thermal object tracking for SAR missions with a drone. The entire process consists of object detection and multiple-target tracking. The You-Only-Look-Once (YOLO) detection model is utilized to detect people in thermal videos. Multiple-target tracking is performed via track initialization, maintenance, and termination. Position measurements in two consecutive frames initialize the track. Tracks are maintained using a Kalman filter. A bounding box gating rule is proposed for the measurement-to-track association. This proposed rule is combined with the statistically nearest neighbor association rule to assign measurements to tracks. The track-to-track association selects the fittest track for a track and fuses them. In the experiments, three videos of three hikers simulating being lost in the mountains were captured using a thermal imaging camera on a drone. Capturing was assumed under difficult conditions; the objects are close or occluded, and the drone flies arbitrarily in horizontal and vertical directions. Robust tracking results were obtained in terms of average total track life and average track purity, whereas the average mean track life was shortened in harsh searching environments.

Generalised covariance intersection‐Gamma Gaussian Inverse Wishart‐Poisson multi‐Bernoulli Mixture: An intelligent multiple extended target tracking scheme for mobile aquaculture sensor networks

AbstractPoisson multi‐Bernoulli Mixture (PMBM) filter has been known as an available or practical point and multiple extended target tracking (METT) method. The authors present an improved PMBM filter with adaptive detection probability and adaptive newborn distributions, accompanying with an associated distributed fusion strategy for the tracking extended multiple targets. First, the augmented state of unknown and changing target detection probability is assumed as Gamma (GAM) distribution. Second, extended states are described by Inverse Wishart (IW) distribution based on this augmented state, accompanying with dynamic states presented by Gaussian distribution. And then, an adaptive newborn distribution is adopted to describe the newborn targets appearing arbitrarily. Consequently, the closed‐form solutions of the proposed filter can be derived by approximating the intensity of newborn and potential targets to the Gamma Gaussian Inverse Wishart (GGIW) form. Moreover, the fused means that Generalised Covariance Intersection (GCI) is performed in such a large‐scale aquaculture sensor network. Experiments are presented to verify the availability of the GCI‐GGIW‐PMBM method, and comparisons with other METT filters also demonstrate that tracking behaviours are improved largely.