Narrow Results

Aiming at the poor anti-occlusion ability of target feature extraction in the process of tennis motion video occlusion target tracking, and the matching problem between target feature map and multiple moving targets, a tennis motion video occlusion target tracking method based on attention mechanism to maximize overlap is proposed. VGG16 neural network is used to stack the features extracted from the convolution layer through a filter to obtain key candidate moving target features. The temporal and spatial attention mechanisms are used to calculate the weight of the extracted candidate moving target features to improve the anti-occlusion ability of moving target feature extraction. The weighted moving target features are input into the support vector machine classifier. Aiming at the defect that SVM can only deal with vector data, a structured support vector machine (SSVM) is proposed. Through the SSVM, the overlap rate between the candidate moving target features and the actual moving target is output. The estimated position of the current video sequence frame is obtained according to the maximum overlap. Sample the frame target in the video sequence, and realize the occlusion target tracking in tennis motion video by cycling the above process. The experimental results show that the accuracy of feature extraction of this method is about 90%. After adding the weight distribution mechanism, the convergence speed is fast, and the mean square error is about 0.2, indicating that the method has a high accuracy of feature extraction. In the detection of overlap rate, the highest overlap rate is 0.912, indicating that the method has a high accuracy of overlap rate. In addition, in the tennis motion video occlusion target tracking and detection, it can accurately locate the target position, and surround all parts of the target. There will be no redundant blank in the target box, which fully reflects the accuracy and robustness of tennis motion video occlusion target tracking.

Having a better motion model in the state estimator is one way to improve target tracking performance. Since the motion model of the target is not known a priori, either robust modeling techniques or adaptive modeling techniques are required. The neural extended Kalman filter is a technique that learns unmodeled dynamics while performing state estimation in the feedback loop of a control system. This coupled system performs the standard estimation of the states of the plant while estimating a function to approximate the difference between the given state-coupling function model and the behavior of the true plant dynamics. At each sample step, this new model is added to the existing model to improve the state estimate. The neural extended Kalman filter has also been investigated as a target tracking estimation routine. Implementation issues for this adaptive modeling technique, including neural network training parameters, were investigated and an analysis was made of the quality of performance that the technique can have for tracking maneuvering targets.

In dealing with the problems of nonlinear and non-Gaussian systems, the Kalman particle filter (KPF) algorithm has been widely used for the aerial and underwater target tracking system, which is vulnerable to environmental interference in recent years, especially in military fields. However, in most of the target tracking research, the indexes used by researchers to evaluate tracking algorithms are usually limited to the state difference or the root mean square error in the tracking process, and the overall tracking effect is only evaluated by the tracking time. These indexes cannot be evaluated for the algorithm that does not have obvious error optimization performance in the tracking process but actually improves the overall tracking effect. Therefore, this paper proposes to evaluate the tracking effect of different tracking algorithms by using the mean square error of relative points in the plane control network of engineering survey. Simulation results show that the proposed performance index combined with the tracking sampling time and the mean square error of tracking position can effectively evaluate the local and global tracking performance of different algorithms.

Sensor networks that can support time-critical operations pose challenging problems for tracking events of interest. We propose an architecture for a sensor network that autonomously adapts in real-time to data fusion requirements so as not to miss events of interest and provides accurate real-time mobile target tracking. In the proposed architecture, the sensed data is processed in an abstract space called Information Space and the communication between nodes is modeled as an abstract space called Network Design Space. The two abstract spaces are connected through an interaction interface called InfoNet, that seamlessly translates the messages between the two. The proposed architecture is validated experimentally on a laboratory testbed for multiple scenarios.

By equaling the detection of centroid jamming to the “outlier” detection in the process of tracking with Kalman filter, an approach to target reselection for anti-ship missile against centroid jamming with accurate tracking information is proposed in this paper considering the mutation of the law of motion at the tracking point under centroid jamming. For this approach, an accurate target tracking model is built on the basis of extended Kalman filter (EKF). Using the information collected by radar seeker including distance, velocity and angle, it could achieve the accurate tracking of the target and determine the motion state of the target accurately. On this basis, the orthogonality of innovation in the process of Kalman filtering is utilized to detect any mutation of the motion state of the target, so that the existence of centroid jamming is detected when there is any mutation of the law of motion at the tracking point in the process of chaff centroid jamming. This offers a new solution for remote beyond-visual-range (BVR) anti-ship missile against centroid jamming.

Environment perception is crucial for the development of autonomous driving and advanced driver assistance systems. The cooperative perception using the infrastructure sensors can significantly expand the field of view of on-board sensors and improve the accuracy of target tracking. In this paper, we propose a hybrid vehicular perception system that incorporates both received feature-level information from infrastructure sensors and track-level data from the multi-access edge computing server (MEC-Server). An infrastructure-enhanced multiple-model probability hypothesis density is proposed to handle the feature-level data from heterogeneous infrastructure sensors. The problem of kinematic state estimation is improved with the prior information of the road environment. Furthermore, a generic communication interface between the infrastructure sensor and MEC-Server is designed, which allows the object data to have the same notion of locality through the use of a generic object state model. Simulation results show that the presented algorithm provides higher accuracy and reliability after considering the prior information of the road environment.

Detection of Stretching And Folding (SAF) traits in a time series is still controversial and of great interest. Also, visuo-manual tracking studies did not pay attention to SAF in hand motion trajectories. This research aims to find out the relevance of SAF to the discontinuities in chaotic dynamics of hand motion through target tracking tasks. Specifically, a new method is constructed based on this relation in which SAF can extract accurately trajectories in both time domain and phase space. Consequently, we designed experiments to track sinusoidal and trapezoidal target movements shown on a monitor. In these experiments, fourteen participants were instructed to move the joystick handle by wrist flexion-extension movements. Results confirm intermittency in significant human motor control behavior which results in discontinuities in hand motion trajectories. The relation between SAF and these discontinuities is realized by chaotic and intermittent behaviors of tracking dynamics. Verification of the method’s accuracy is also carried out by taking advantage of the Poincaré section. Our method can provide insight into the dynamical behaviors of chaotic and intermittent systems involving mechanisms in human motor control. It can be applied to general systems with intermittent behavior and other systems with modification.

Automatic target tracking is a challenging task in video surveillance applications. Here, an offline target-tracking system in video sequences using Discrete Wavelet Transform is presented. The proposed algorithm uses co-occurrence features, derived from sub-bands of discrete wavelet transformed sub-blocks, obtained from individual video frames, to identify a seed in the frame. Then, the region-growing algorithm is applied to detect and track the target. The results of the proposed target detection and tracking system in video sequences are found to be satisfactory. The effectiveness of the target-tracking algorithm has been proved as the target gets detected, irrespective of size of the target, perspective view and cluttered environment.

In human–robot interaction developments, detection, tracking and identification of moving objects (DATMO) constitute an important problem. More specifically, in mobile robots this problem becomes harder and more computationally expensive as the environments become dynamic and more densely populated. The problem can be divided into a number of sub-problems, which include the compensation of the robot's motion, measurement clustering, feature extraction, data association, targets' trajectory estimation and finally, target classification. Here, a mobile robot uses 2D laser range data to identify and track moving targets. A Joint Probabilistic Data Association with Interacting Multiple Model (JPDA-IMM) tracking algorithm associates the available laser data to track and provide an estimated state vector of targets' position and velocity. Potential moving objects are initially learned in a supervised manner and later on are autonomously classified in real-time using a trained Fuzzy ART neural network classifier. The recognized targets are fed back to the tracker to further improve the track initiation process. The resulting technique introduces a computationally efficient approach to already existing target-tracking and identification research, which is especially suited for real time application scenarios.

Prediction in real-time image sequences is a key-feature for visual servoing applications. It is used to compensate for the time-delay introduced by the image feature extraction process in the visual feedback loop. In order to track targets in a three-dimensional space in real-time with a robot arm, the target's movement and the robot end-effector's next position are predicted from the previous movements. A modular prediction architecture is presented, which is based on the Kalman filtering principle. The Kalman filter is an optimal stochastic estimation technique which needs an accurate system model and which is particularly sensitive to noise. The performances of this filter diminish with nonlinear systems and with time-varying environments. Therefore, we propose an adaptive Kalman filter using the modular framework of mixture of experts regulated by a gating network. The proposed filter has an adaptive state model to represent the system around its current state as close as possible. Different realizations of these state model adaptive Kalman filters are organized according to the divide-and-conquer principle: they all participate to the global estimation and a neural network mediates their different outputs in an unsupervised manner and tunes their parameters. The performances of the proposed approach are evaluated in terms of precision, capability to estimate and compensate abrupt changes in targets trajectories, as well as to adapt to time-variant parameters. The experiments prove that, without the use of models (e.g. the camera model, kinematic robot model, and system parameters) and without any prior knowledge about the targets movements, the predictions allow to compensate for the time-delay and to reduce the tracking error.

The theory and technology of human–machine coordination and natural interaction have a wide range of application prospect in future smart factories. This paper elaborates on the design and implementation of a body-following wheeled robot system based on Kinect, as well as the use of gesture recognition function to enhance the interactive performance. An improved optical flow method is put forward to obtain the direction and speed of the target movement. The smoothing parameters in traditional optical flow are replaced by variables. The new smoothing parameter is related to the local gradient value. Compared with the traditional optical flow method, it can reflect the status of moving objects more clearly, reduce noise and ensure real-time performance, solving the problem of tracking state oscillation caused by the skeleton node drifts when the target is occluded. The experiment on the wheeled robot confirms that the system can accomplish the tracking task in a preferable way.

In this paper, we present a comprehensive design for a fully functional unmanned rotorcraft system: GremLion. GremLion is a new small-scale unmanned aerial vehicle (UAV) concept using two contra-rotating rotors and one cyclic swash-plate. It can fit within a rucksack and be easily carried by a single person. GremLion is developed with all necessary avionics and a ground control station. It has been employed to participate in the 2012 UAVForge competition. The proposed design of GremLion consists of hardware construction, software development, dynamics modeling and flight control design, as well as mission algorithm investigation. A novel computer-aided technique is presented to optimize the hardware construction of GremLion to realize robust and efficient flight behavior. Based on the above hardware platform, a real-time flight control software and a ground control station (GCS) software have been developed to achieve the onboard processing capability and the ground monitoring capability respectively. A GremLion mathematical model has been derived for hover and near hover flight conditions and identified from experimental data collected in flight tests. We have combined H_∞ technique, a robust and perfect tracking (RPT) approach, and custom-defined flight scheduling to design a comprehensive nonlinear flight control law for GremLion and successfully realized the automatic control which includes take-off, hovering, and a variety of essential flight motions. In addition, advanced mission algorithms have been presented in the paper, including obstacle detection and avoidance, as well as target following. Both ground and flight experiments of the complete system have been conducted including autonomous hovering, waypoint flight, etc. The test results have been presented in this paper to verify the proposed design methodology.

This paper considers the problem of localization and circumnavigation of a slowly drifting target with an unknown speed by a group of autonomous agents while they form a regular polygon at a known distance from the target. The goal is achieved in a distributed way where each of the agents coordinates its motion knowing its own position and either the bearing angle of the target or the distance to the target, and the position of one of its neighbors. First, we solve the problem for the case where the target is stationary and propose a two-stage control law that forces the agents to move on a circular trajectory around the target and form a regular polygon formation. Then, we consider the case where the target is undergoing a slow but possibly persistent movement. Later, we consider the case where only one of the agents know the desired distance from the target. In the end, the case in which only a subset of agents can measure either the bearing or the distance to the target is considered. The performance of the controllers proposed is verified analytically, through simulations, and in an experimental setup.

This paper focuses on the development of control and guidance laws for quadrotor Unmanned Aerial Vehicles (UAVs) to track maneuvering ground targets. Proportional Derivative (PD) control law is a popular choice to be used as a tracking controller for quadrotors, but it is often inefficient due to practical acceleration constraints and a number of parameters that need to be tuned. The paper proposes a Proportional Navigation (PN)-based switching strategy to address the problem of mobile target tracking. The experiments and numerical simulations performed using nonmaneuvering and maneuvering targets show that the proposed PN-based switching strategy not only carries out effective tracking but also results into smaller oscillations and errors when compared to the widely used PD tracking method. The proposed PN-based switching strategy presents an important question with regard to when the switching should happen that would minimize the positional error between the UAV and the target. An optimal switching strategy, which is based on the analytical solutions of the PN and PD methods, is proposed. The numerical simulations not only validate the theoretical results with regard to the optimality of the proposed method for both nonmaneuvering and maneuvering targets but also demonstrate that the proposed method is robust to measurement noise.

The interest in autonomous marine vessels has been continuously growing in the recent years. Most platforms of the autonomous surface watercraft involve traditional mono- or multi-hulls. Advanced marine vehicle concepts, such as hydrofoils, can provide high-speed and high seakeeping capabilities. In this study, a modeling effort is initiated for a small autonomous hydrofoil boat intended for intercepting operations. A 3-DOF model, including surge, sway and yaw, is applied for simulating maneuvering motions of the boat in the foilborne state. Forces generated by the propulsor, rudder and struts are accounted for in the simulations of the horizontal-plane boat dynamics. Two scenarios of a hydrofoil boat pursuing a moving target are investigated. In the pure pursuit, the interceptor always attempts to aim at the target and uses full thrust to quickly reach the target at a high speed. In the constant-bearing scenario, the interceptor approaches the target with diminishing speed trying to achieve a rendezvous. The presented models and results can help engineers to design more effective control methods for fast boats intended for intercepting operations.

Unmanned aerial vehicles (UAVs) are experiencing a rapid expansion in their applications across various domains, including goods delivery, video capturing, and traffic control. The crucial aspect for UAVs to execute successful target tracking and obstacle avoidance maneuvers lies in the accuracy of their path planning operations. This research paper aims to contribute to the existing body of knowledge by presenting a novel model that incorporates acceleration control, accounting for changing variables such as UAV velocity and altitude, while also incorporating vehicle dynamics. To enhance the realism of the model, we include drag force as a factor. In this study, we focus on exploring the potential of deep reinforcement learning (DRL), specifically the deep deterministic policy gradient (DDPG) algorithm, for modeling a 3D continuous environment with a continuous set of actions. In order to improve the UAV’s performance in executing target tracking and obstacle avoidance maneuvers, we propose an innovative reward function based on the inner product. The training results show that the UAV successfully learns to perform the aforementioned tasks. Also, simulation results demonstrate the superior performance of the proposed UAV modeling and reward function compared to existing works.

In this paper, the problem of searching and tracking uncooperative and unidentified mobile ground target using a quadcopter unmanned aerial vehicle (QUAV) is investigated. The proposed strategy is an Image-Based Visual Servoing (IBVS) approach, combined with the virtual camera concept and robust control. This strategy makes use of the rough prior information of the target, in contrast with existing strategies, which will alleviate the problem of altitude estimation noise and enhance the overall accuracy. Hence, a new vision-based sliding mode controller (SMC) is designed to control the quadcopter taking into account the flight phase’s heterogeneity, the external disturbances and parameters uncertainties as well as the target maneuverability. In order to get a better insight about the SMC tuning and adjustment, three different reaching laws are evaluated and compared. The proposed controller allows an automatic execution of the flight strategy whilst the searching phase relies on the Camera Coverage Area (CCA) technique. The vision-based technique allows an automatic QUAV altitude tuning for optimal target observation and tracking. Another contribution of this work is the fact that the designed controller validity and stability overspan the entire scenario to reach the universal and to smoothen out surges generated by control switching. Numerical simulations are conducted to compare the proposed SMC controllers and validate the effectiveness of the whole strategy.

Target detection and tracking represent key challenges facing miniature fixed-wing unmanned aerial vehicles (UAVs), particularly at high cruising speeds. Therefore, this paper proposes a vision-based target detection and tracking algorithm that systematically couples two mainstream methods, namely, you only look once (YOLO) and kernel correlation filter (KCF) algorithms. This combination enables small fixed-wing UAVs to achieve reliable target detection and rapid target tracking. A customized vision-guidance module is constructed to implement this algorithm, and a dual-thread execution mechanism is developed to ensure that the computational resources are used effectively. A miniature fixed-wing UAV experimental platform is also constructed and evaluated. Flight experiments are performed, and the results demonstrate that the developed algorithm can achieve satisfactory detection and tracking accuracy for stationary and moving ground targets in complex environments.

In this paper, the Quadrature Kalman filter (QKF) algorithm of nonlinear systems is studied in-depth, with its merits and drawbacks being analyzed. A suboptimal fading QKF algorithm (SFQKF) based on Strong Tacking Filter(STF) is also proposed to adjust the covariance matrix of state prediction error, the covariance matrix of prediction error, and the cross covariance matrix between the state prediction error and the measured prediction error in real time through the time-varied suboptimum fading factor, which can adjust the gain matrix of filter in real time. Moreover, the derivative process of suboptimal fading factor is given. The mechanism analysis and emulation experiment of this algorithm show that SFQKF algorithm, which inherits the excellent performance of Strong Tracking Filter (STF), can overcome the defects of QKF algorithm and have stronger ability to track the states with abrupt changes. Compared with QKF algorithm, the stability of FQKF algorithm is improved by 14.9%, and the amount of calculations required is moderate.

This article presents an illustration of moving track vehicles, builds an echo model of scatter points on the track, and analyzes impacts of various motion trends on radar echo. Moreover, it conducts simulation study in vehicles with different attitude angles based on micro-motion feature. Experimental results validate the effectiveness of this model.