Target Tracking Techniques Research Articles

UAV target tracking method based on global feature interaction and anchor-frame-free perceptual feature modulation.

Target tracking techniques in the UAV perspective utilize UAV cameras to capture video streams and identify and track specific targets in real-time. Deep learning UAV target tracking methods based on the Siamese family have achieved significant results but still face challenges regarding accuracy and speed compatibility. In this study, in order to refine the feature representation and reduce the computational effort to improve the efficiency of the tracker, we perform feature fusion in deep inter-correlation operations and introduce a global attention mechanism to enhance the model's field of view range and feature refinement capability to improve the tracking performance for small targets. In addition, we design an anchor-free frame-aware feature modulation mechanism to reduce computation and generate high-quality anchors while optimizing the target frame refinement computation to improve the adaptability to target deformation motion. Comparison experiments with several popular algorithms on UAV tracking datasets, such as UAV123@10fps, UAV20L, and DTB70, show that the algorithm balances speed and accuracy. In order to verify the reliability of the algorithm, we built a physical experimental environment on the Jetson Orin Nano platform. We realized a real-time processing speed of 30 frames per second.

Using posture recognition algorithms based on machine learning to identify senior health

Faced with the situation that the elderly people at home have dangerous behaviors, the study explores various aspects of motion target detection, real-time target tracking and behavioral pose recognition and classification, using behavioral poses in videos as samples. To tackle the challenges in detecting motion targets, a target detection method based on Gaussian mixture model (GMM) and four frame difference method is proposed; A tracking technique incorporating Kalman filter (KF) is investigated to trail the behavioral changes of the elderly in actual time. A seven-layer convolutional neural network (CNN) is constructed to face the problem of inaccurate behavioral pose recognition. Through relevant experimental analyses, the outcomes show that the increased GMM detection way has a complete profile and the accuracy is significantly improved. The KF target tracking technique can trail the object trajectory in actual time and steadily, with the smallest trailing error value of 0.19. The classification accuracy of the CNN pose recognition model is 95.87%, and the pose classification time is 27 seconds. Its performance is superior to the mean shift algorithm, particle filter algorithm, and Cam Shift algorithm in all aspects. When applied in practice, it can accurately identify whether the elderly’s behavior is abnormal and ensure their daily health.

A novel data-driven framework for enhancing the consistency of deposition contours and mechanical properties in metal additive manufacturing

The accuracy and quality of part formation are crucial considerations. However, the laser directed energy deposition (L-DED) process often leads to irregular changes in deposition contours and mechanical properties across parts due to complex flow fields and temperature variations. Hence, to ensure the forming accuracy and quality, it is necessary to achieve precise monitoring and appropriate parameter adjustments during the processing. In this study, a machine vision method for real-time monitoring is proposed, which combines target tracking and image processing techniques to achieve accurate recognition of deposition contours under noisy conditions. Through comparative verification, the measurement accuracy reaches as high as 98.98 %. Leveraging the monitoring information, a bidirectional prediction neural network is proposed to accomplish layer-by-layer forward prediction of layer height. Meanwhile, inverse prediction is employed to determine the processing parameters required for achieving the desired layer height, facilitating the optimization of the deposition contours. It was found that as the processing parameters were adjusted layer-by-layer to achieve consistent deposition contours, there was also a tendency towards consistent changes in microstructure and mechanical properties. The standard deviations of primary dendrite arm spacing (PDAS) and ultimate tensile strength (UTS) at different positions decrease by over 52.2 % and 61.4 %, respectively. This study reveals the consistent patterns of variation in deposition contours and mechanical properties under data-driven variable parameter processing, laying an important foundation for future exploration of the complex process-structure-performance (PSP) relationship in L-DED.

Vortex-induced vibration dynamic characteristics monitoring with AI-based target object detection for long-span bridges

This research proposes an innovative framework for monitoring bridge Vortex-induced Vibration (VIV) dynamic characteristics utilizing AI-based machine-learning target object detection and tracking techniques with keypoint detection. Two camera calibration methods are implemented for cases with and without intrinsic and extrinsic camera setup information based on homography matrix conversion and distance-based conversion. The framework is verified on a video recording of a real-bridge VIV event and a simulation animation of bridge VIV with a peak/trough-based statistical method for calculating VIV frequency and amplitude. Accurate detection is achieved in both cases with short video durations. The framework demonstrates great potential for real-time bridge VIV monitoring, requiring minimal camera calibration and a straightforward device setup. It offers reliable and accurate results while remaining cost-effective.

Target tracking techniques for multi-robot systems: Review on the state-of-the-art

The field of target tracking has long been a focal point in robotics research, gaining particular prominence in the realm of multi-robot systems. Target tracking has extensive applications in diverse areas such as surveillance, search and rescue operations, and environmental detection. This paper presents an exhaustive review of the latest target tracking methodologies employed in multi-robot systems, with an emphasis on scenarios featuring both abundant and limited information availability. The review particularly addresses the challenges posed by conditions including target occlusion (partial or total), uncertainties in robots positional data, and limitations in sensory and communicative functions. This scholarly piece synthesizes literature sourced from IEEE and Springer publications, spanning the period from 2019 to 2023. This temporal scope ensures a focus on contemporary and pertinent research outcomes. The findings of this paper indicate a significant evolution in the sophistication of target tracking techniques, particularly in complex operational scenarios. These advancements signify a notable maturation in the capabilities of multi-robot systems to efficiently handle intricate and varied tasks.

Improved single target identification tracking algorithm based on IPSO-BP neural network

Abstract Driven by deep learning techniques in recent years, single target recognition and tracking techniques have developed significantly, but face challenges of real-time and accuracy. In this study, an improved IPSO-BP network is formed by optimizing three critical aspects of the IPSO algorithm: adjusting the inertia weight calculation formula, improving the learning factor, and creating a new iterative formula for particle updating, which in turn is combined with a BP neural network. After iterative training, this paper constructs a single target recognition tracking algorithm with higher efficiency. The Algorithm’s performance is comprehensively tested through experimental simulation in terms of real-time, accuracy and stability. The results show that the improved Algorithm can achieve a frame rate (FPS) of up to 31 in single target recognition and tracking. The IOU value is as high as about 83% in some tests. The tracking success rate in different scenarios averages approximately 98.50%, the position error is controlled within 0.7 m, and the speed error averages 2.75 m/s. This improved IPSO-BP neural network effectively solves the problems of the current technology in the areas of real-time and accuracy, showing high stability and accuracy.

Modeling and tracking control of underwater unmanned underwater vehicles for deep-sea marine environment detection

Abstract This paper focuses on enhancing target tracking techniques for unmanned underwater vehicles (UUVs) used for deep-sea exploration, especially for the latest advances in digital technology. The article begins with an innovative UUV target tracking algorithm improvement based on the correlation filtering algorithm. Introducing the spatio-temporal regularization method and penalty function optimizes the filter parameters and enhances the algorithm’s performance. Then, a new objective function is constructed by combining the optimal feature weights and baseline function calculated by the dynamic feature weighting method, which effectively excludes the influence of interference peaks in the algorithm. In addition, by applying the Gaussian correlation method, the article further improves the algorithm’s robustness. When the obstacle avoidance ability of the algorithm is analyzed, it is found that the control error of the obstacle avoidance speed in both X and Y directions is less than 0.1 m/s, and the difference between the actual angle and the desired angle of the attitude angle is less than 0.5 degrees. This indicates that the algorithm can effectively avoid all obstacles and accurately track the target. The target tracking experiments show that the tracking performance scores of the algorithm are generally higher than 9, and the tracking accuracy is more than 95%, which is significantly better than other existing algorithms. The optimized tracking algorithm proposed in this paper not only provides essential technical support for the development of unmanned underwater vehicle technology, but also provides valuable reference for the future application and practice of the algorithm.

An Analytical Study on Improving Target Tracking Techniques in Wireless Sensor Networks Using Deep Learning and Energy Efficiency Models

The work of Wireless Sensor Network (WSN) in goal tracking presentations, particularly in military, environmental, and surveillance operations, has garnered significant attention. In order to progress the accuracy, efficiency, and energy protection of these networks, a diversity of methods have been invented to resolve challenges such as real-time monitoring, multiple-target tracking, and resource optimization. This study examines a broad spectrum of algorithms and frameworks, such as energy-efficient methods, clustering methods, augmented state-based algorithms, and deep learning-based models, to address target tracking issues in WSNs. The outcomes emphasize the progress in the field, with a particular emphasis on enhancing tracking accuracy and reducing energy consumption, ranging from adaptive clustering methods to multi-sensor deep learning outlines. Recurrent neural networks (RNN), long short-term memory (LSTM), and prognostic models are instrumental in improving target detection, localization, and tracking. The examination also delves into the placement of WSN in a diversity of environments, including battlefields, intelligent transportation systems, and quantized areas, providing valuable visions into future trends and research orientations.

Camera-Based Local and Global Target Detection, Tracking, and Localization Techniques for UAVs

Multiple-object detection, localization, and tracking are desirable in many areas and applications, as the field of deep learning has developed and has drawn the attention of academics in computer vision, having a plethora of networks now achieving excellent accuracy in detecting multiple objects in an image. Tracking and localizing objects still remain difficult processes which require significant effort. This work describes an optical camera-based target detection, tracking, and localization solution for Unmanned Aerial Vehicles (UAVs). Based on the well-known network YOLOv4, a custom object detection model was developed and its performance was compared to YOLOv4-Tiny, YOLOv4-608, and YOLOv7-Tiny. The target tracking algorithm we use is based on Deep SORT, providing cutting-edge tracking. The proposed localization approach can accurately determine the position of ground targets identified by the custom object detection model. Moreover, an implementation of a global tracker using localization information from up to four UAV cameras at a time. Finally, a guiding approach is described, which is responsible for providing real-time movement commands for the UAV to follow and cover a designated target. The complete system was evaluated in Gazebo with up to four UAVs utilizing Software-In-The-Loop (SITL) simulation.

Set-membership vs. stochastic approaches for target localization with UAVs

Searching and tracking mobile targets rely most often on modeling the uncertainty and various perturbations by stochastic processes. The detection and location of the targets are performed with Bayesian estimation, which reliability and resulting performance are deeply linked to the adequacy of the stochastic models. An alternative approach limits the representation of these perturbations by defining the bounds within which they can vary. Set-membership estimation techniques have been developed to handle this representation. This paper compares the performance of set-membership and stochastic Bayesian estimation techniques for target search and tracking for scenarios integrating false alarms. For this purpose, estimation schemes are presented for each approach. The ability of estimators to find real targets and not to be deceived by false targets or imperfect sensors are compared in simulations.

Gaussian processes based extended target tracking in polar coordinates with input uncertainty

Gaussian processes (GP) based extended target tracking (ETT) technique has attracted the attention of scientists since it can accurately estimate the kinematic states as well as the contour states of irregular-shape targets. Most traditional GP methods consider the ETT problem with a linear measurement model in Cartesian coordinates and ignore the input uncertainty of GP for simplicity. In many applications, however, measurements are generated with high-resolution sensors in polar coordinates. More importantly, the ignorance of input uncertainty in GP will exacerbate the tracking performance in these cases. In order to track an irregular-shape extended target in polar coordinates with input uncertainty, an improved GP-based probabilistic data association (IGP-PDA) algorithm is developed that includes the following enhancements: Firstly, a nonlinear measurement model is used and the unbiased converted measurement technique is invoked for the ETT problem. Secondly, the analytical form of statistical properties of the input uncertainty due to measurements noise and predicted error is derived. Thirdly, an IGP-PDA taking into account measurement origin uncertainty as well as input uncertainty is proposed, and three sub-optimal implementations are provided. Last, the posterior Cramer-Rao lower bound of ETT with input uncertainty is derived. Simulation results verify the effectiveness of the proposed method.

Multitarget Tracking Algorithm in Intelligent Analysis of Football Movement Training Stance

In recent years, with the continuous development of computer technology, deep learning has been widely applied to computer vision tasks and has achieved great success in areas such as visual detection and tracking. On this basis, making deep learning techniques truly accessible to people becomes the next objective. Target detection and tracking in football gesture training is a quite challenging task with great practical and commercial value. In traditional football training methods, target trajectories are often extracted by means of a recording chip carried by the player. However, the cost of this method is high and it is difficult to replicate in amateur stadiums. Some studies have also used only cameras to process targets in football videos. However, due to the similarity in appearance and frequent occlusion of targets in football videos, these methods often only segment targets such as players and balls in the image but do not allow them to be tracked. Target tracking techniques are of great importance in football training and are the basis for tasks such as player training analysis and match strategy development. In recent years, many excellent algorithms have emerged in the field of target tracking, mainly in the categories of correlation filtering and deep learning, but none of them are able to achieve high accuracy in player tracking for football training videos. After all, the problem of locating clips of interest to athletes from a full-length video is a pressing one. Traditional machine learning-based approaches to sports event detection have poor accuracy and are limited in the types of events they can detect. These traditional methods often rely on auxiliary information such as audio commentary and relevant text, which are less stable than video. In recent years, deep learning-based methods have made great progress in the detection of single-player video events and actions, but less so in the detection of sports video events. As a result, there are few sports video datasets that can be used for deep learning training. Based on research in computer vision and deep learning, this paper designs a multitarget tracking system for football training. To be specific, this algorithm uses multiple cameras for image acquisition in the stadium in order to accurately track multiple targets in the stadium over time. Furthermore, the framework for a single camera multitarget tracking approach has been designed based on deep learning-based visual detection methods and correlation filter-based tracking methods. This framework focuses on using data correlation algorithms to fuse the results of detectors and trackers so that multiple targets can be tracked accurately in a single camera. To sum up, this research allows for robust and real-time long-term accurate tracking of targets in football training videos through multitarget tracking algorithms and the intercorrection of multiple camera systems.

Design and Implementation of Sensor Platform for UAV-Based Target Tracking and Obstacle Avoidance

Small-scale unmanned aerial vehicles are being deployed in urban areas for missions such as ground target tracking, crime scene monitoring, and traffic management. Aerial vehicles deployed in such cluttered environments are required to have robust autonomous navigation with both target tracking and obstacle avoidance capabilities. To this end, this work presents a simple-to-design but effective steerable sensor platform and its implementation techniques for both obstacle avoidance and target tracking. The proposed platform is a 2-axis gimbal system capable of roll and pitch/yaw. The mathematical model that governs the dynamics of this platform is developed. The performance of the platform is validated through a software-in-the-loop simulation. The simulation results show that the platform can be effectively steered to all regions of interest except backward. With its design layout and mount location, the platform can engage sensors for obstacle avoidance and target tracking as per requirements. Moreover, steering the platform in any direction does not induce aerodynamic instability on the unmanned aerial vehicle in mission.

Motion Modeling and State Estimation in Range-Squared Coordinate

In some target tracking applications, angle measurements are absent due to sensor characteristics, which makes Cartesian states unobservable for a single stationary sensor. When multi-sensor target tracking techniques with only range or Doppler measurements are applied to produce Cartesian state estimates, their performance and efficiency can be benefited from accurate range and Doppler estimates obtained from single sensors. This paper investigates the range and Doppler estimation problem based on motion modeling in range-squared coordinate considering range-squared as a generic coordinate. A pseudostate vector consisting of range squared and its derivatives is defined to completely summarize the dynamic of the Cartesian nearly constant velocity (NCV) motion and it is found time evolution of the pseudostate is linear based on the methodology of state space representation. To estimate the pseudostate from range-Doppler or range-only measurements, a recursive filter is developed using the decorrelated unbiased converted measurement (DUCM) method. Finally, the pseudostate estimates in the range-squared coordinate are converted into the range coordinate using the scaled unscented transformation (SUT) method to produce range and Doppler estimates. Although the conversion is nonlinear, it is performed outside the filtering recursions without propagating nonlinearity approximation errors in the filtering recursions. Numerical experiments with various conditions of measurements demonstrate the superior performance benefits from the proposed accurate linear state equation and estimation method.

Distributed trust‐based unscented Kalman filter for non‐linear state estimation under cyber‐attacks: The application of manoeuvring target tracking over wireless sensor networks

This paper is concerned with secure state estimation of non-linear systems under malicious cyber-attacks. The application of target tracking over a wireless sensor network is investigated. The existence of rotational manoeuvre in the target movement introduces non-linear behaviour in the dynamic model of the system. Moreover, in wireless sensor networks under cyber-attacks, erroneous information is spread in the whole network by imperilling some nodes and consequently their neighbours. Thus, they can deteriorate the performance of tracking. Despite the development of target tracking techniques in wireless sensor networks, the problem of rotational manoeuvring target tracking under cyber-attacks is still challenging. To deal with the model non-linearity due to target rotational manoeuvres, an unscented Kalman filter is employed to estimate the target state variables consisting of the position and velocity. A diffusion-based distributed unscented Kalman filtering combined with a trust-based scheme is applied to ensure robustness against the cyber-attacks in manoeuvring target tracking applications over a wireless sensor network with secured nodes. Simulation results demonstrate the effectiveness of the proposed strategy in terms of tracking accuracy, while random attacks, false data injection attacks, and replay attacks are considered.

Mobile target localization and tracking techniques in harsh environment utilizing adaptive multi‐modal data fusion

Multi-source cooperative positioning systems relying on federated filtering have become attractive development directions of navigation strategy in real-time localization and tracking under challenging urban environment. However, local Kalman filters of traditional federated filtering may result in divergence when the system model or the measurements is inaccurate, and the fixed information distribution coefficient in federated filter cannot adaptively reflect the performance of each local filter. To improve the precision and robustness of the integrated navigation system, a novel adaptive federated strong tracking Kalman filter with dynamic fading factor mechanism for multi-sensor information fusion is proposed. Through iterative computation of the fading factor and updating the adaptive weight coefficients, strong tracking filter becomes robust to the uncertainty of system model. Meanwhile, an effective adaptive information distribution estimation algorithm based on the predicted residuals is constructed to balance the contributions of the kinematic model information and measurements on the state estimates. To ensure the stability of filtering, a simplified fusion strategy is established to solve the singular problem of the global covariance matrix of estimation error. Theoretical analysis and simulation results demonstrate the validity of the proposed approach in improving the accuracy and robustness of the integrated navigation and positioning systems. The proposed integrated multi-modal cooperative navigation and positioning algorithm will be of great significance to the implementation of real-time mobile target localization and tracking in harsh environment or dead zone.

The Technology of Moving Target Detection and Tracking Based on Sequential SAR Image

Although the multi-channel SAR-GMTI technology can be used in ground moving target detection, but only can get trace point of the ground moving target, the ground moving target tracking and track reconstruction can’t be achieved. This article presents a new method of ground moving target detection and tracking which is based on sequence SAR images. Firstly, the moving target detection and velocity estimation can be realized using the method of multi-channel processing. Secondly, the location of moving target can be estimated accurately using a high accuracy target location method based on knowledge-aid. The method uses the matching and target tracking technique based on sequential SAR images to realize the tracking and track reconstruction of moving target. Next, the image reference locking technology is used to achieve accurate image registration for a series of sequential SAR images obtained. Finally, the moving target can be tracked. The results of measured data shows the validity of the presented method.

Implementation Of underwater target tracking techniques for Gaussian and non-Gaussian environments

There are several techniques implemented, in an underwater target tracking environment, for the nonlinear dynamic systems in Gaussian and non-Gaussian environments. It is assumed with non-Gaussian distribution to make the problem part of the non-Gaussian distribution, and is measured in terms of calculations of plenty of scenarios simulated to validate the potential of the sub-optimal filter.This research is further carried out by considering two categories of non-Gaussian noises i.e.a mixture of Gaussian noises and shot noise. To evaluate tracking in Gaussian and non-Gaussian noises, the suboptimal filters, Extended Kalman filter, and Unscented Kalman filter (UKF algorithms are considered. Gaussian noise is a statistical noise having probability density function equal to the normal distribution function. The suboptimal filters, Extended Kalman filter, and Unscented Kalman filter (UKF) algorithms are considered to evaluate tracking in Gaussian and non-Gaussian noises. To make further evaluation of the above said algorithms, they are compared with theoretical Cramer-Rao lower bound. The efficiency of UKF is in terms of percentage of non-Gaussian noise corrupted measurements, for which solution is obtained within a short time. The application of Monte-Carlo method at this simulations trapped accurate results.

REAL-TIME AUTOMATED CONTOUR BASED MOTION TRACKING USING A SINGLE-CAMERA FOR UPPER LIMB ANGULAR MOTION MEASUREMENT

Upper limb angular motion measurement is particularly a subject of interest in the field of rehabilitation. It is commonly measured manually by physiotherapist via goniometer to monitor stroke patient’s progress. However, manual measurement has many inherent drawbacks such as the need to have both patient and physiotherapist to be at the same place. In this paper, an automated real-time single-camera upper limb angular motion measuring system is proposed for home-based rehabilitation to aid physiotherapist to monitor patient’s progress. The measurement of concern are angle measurement of elbow extension, elbow flexion, wrist flexion and wrist extension. To do this, we propose a method that utilized predefined coordinate points extracted from the contours of the object named as contour based motion tracking. The proposed method overcome problems of prior target tracking techniques such as Kalman filter, Optical flow and Cam-shift. The proposed method includes skin region segmentation and arm modelling for motion tracking. Prior skin segmentation techniques suffer from fixed threshold value set by the user. Therefore, we introduce dynamic threshold based on the lower and upper threshold boundary of isolated skin regions from the background. To ensure the reliability of our skin segmentation method, we compare them with four different related algorithms. The result shows that our skin segmentation method outperformed the prior method with 93% detection accuracy. Following the segmentation, we model the arm motion tracking by formulating mathematical equation of various points and motion velocities to track the arm. We then model the wrist and elbow position to estimate arm angular motion. The method is put together and tested with real human subject and other test settings. The result shows that our proposed method able to produce an accurate and reliable reading of +-1.25 average range of error from actual physiotherapist reading.

Face-Based Mobile Target Tracking Technique in Wireless Sensor Network

A prediction-based method is presented to track mobile object and its location in a sensor network area. In recent years, energy consumption and high accuracy target tracking have been a challenge in wireless sensor network. A number of applications have been used to reduce energy depletion by involving only a few number of sensor nodes to contribute in communication, sensing for target tracking, and transaction. In this study, the Face-based Target Tracking Technique (FTTT) is used to minimise energy depletion to extend the lifetime of sensor node as well as to track object accurately. FTTT combines prediction algorithm with face routing, which can produce accurate detection. At the beginning, sensor node in the border detects the object and elects Triangular sensor Nodes (TN) in the face structure which are the nearest to the object. The process is then continued with TNs tracking the Moving Object (MO) and predicting its next position by face routing structure. The next set of face-based structures will activate TN and offer continuous tracking of the MO. The simulation of the proposed technique are evaluated against the existing approach in terms of object tracking, object detection, speed monitoring as well as energy consumption. The results of FTTT show high accuracy of object detection with less energy consumption.