Multiple Model Recursive Gaussian Process for Robust Target Tracking

Effective and efficient low-altitude flying target tracking in the field of visual tracking is challenging due to factors such as background interference, a small target imaging area, scale changes, and in-plane/out-of-plane rotation. Fast compressive tracking is an effective algorithm that combines compressive sensing theory and the naive Bayes classifier to track targets in real-time. Since the target motion information is not used in the tracking process and a fixed learning rate is adopted, the target may be lost during tracking, especially when the background interference is considerable or when in-plane/out-of-plane rotation exists. To solve this problem, first, target motion information was introduced to reduce the search area for predicting the target position. Then, the confidence calculation was optimized by comprehensively considering the posterior probability of the candidate region and the positive sample membership value. Finally, the learning rate was dynamically adjusted according to the target velocity and optimized confidence. Experimental results verified that the proposed method could effectively improve the efficiency, accuracy, and robustness of target tracking.

In order to enhance the robustness of IR fast small target tracking, a novel mean shift tracking algorithm using improved similarity measure of is proposed. Firstly, problems of local background interfering in original mean shift algorithm for tracking fast motion small target is analyzed, and the reasons is located in the Bhattacharyya coefficient similarity measure expression for all gray weights of components are same, which cannot reflect the advantage contribution of the small target’s gray component in the process of measuring similarity, causing serious interference of the background in the tracking process, leaving the algorithm converging easily. Therefore, to solve this problem, the improvements Bhattacharyya coefficient similarity measure with the local background information fused is proposed. Then, shift vector is deduced in the framework of mean shift by regarding Bhattacharyya coefficients as the similarity measure.In shifting process, the robustness of the small target tracking is improved effectively according to target gray level of large membership degree with high shift weight, and vice versa with low shift weight, which the background interference is suppressed to some extent. In sake of verifying the performance of the proposed algorithm, the classical mean shift algorithm and the algorithm of this paper is used in the target tracking simulation experiment, as well as the infrared image sequences containing the small fast targets of uncooled infrared camera is used. Finally the experimental result indicates that the performance of tracking the small fast target in IR images is robust.

Target detection and tracking is a key technology in intelligent monitoring system. However, it is still difficult to design a robust, accurate and real-time target detection and tracking algorithm. Taking target Tracking as the research focus, this paper focuses on one of the current mainstream algorithms, namely, TLD (tracking-learning-detection) target Tracking algorithm. Based on the theory of TLD target tracking and corner detection, a new target tracking method was proposed to solve the existing problems. A classifier is built online to track and learn the target in real time, and a target contour detector is added in the system to complement the random forest detector to solve the problem of tracking failure. Experimental results show that the proposed algorithm can effectively overcome the influence of fast change or depth occlusion, and improve the tracking efficiency of moving targets.

Object tracking is an important technology in video surveillance. The main approach is Mean Shift algorithm and its improved version. Studies show that the traditional Mean Shift algorithm adopts a fixed searching window in the tracking process, which cannot adjust the template adaptively. The improved algorithm, CamShift, overcomes this problem with an adaptively changing searching window. However, these algorithms are both based on color tracking, which requires that the colors of the foreground targets are unique. If the color of the target is similar to the color of the background, tracking errors will occur or tracking targets will be lost. In this study, we developed an adaptive gradient enhanced texture based tracking algorithm for traffic monitoring applications. This algorithm combines the characteristics of the color and texture of objects. The algorithm builds a joint histogram template of color and texture for targeting, which solves the problems of tracking targets losing when the color of the object is similar to the color of the background. The experiments show that the algorithm can improve the accuracy and robustness of object tracking.

Recently, kernelized correlation Filter-based trackers have aroused the interest of many researchers and achieved good results in the field of tracking. However, the current tracking model based on kernelized correlation filters can not deal with the changes of the target appearance and scale effectively. Therefore, in this paper, we intend to solve these two problems and improve the robustness of the tracking method. Two robust loss functions are offered to address the problem of overfitting in the training process. In addition, a separate scale filter is learned for scale estimation during the tracking process. A large number of experimental results in a variety of complex environments demonstrate that the robustness of visual tracking based on kernelized correlation filters has been greatly improved by our robust collaborative regularization correlation model.

Joint trajectory tracking and recognition based on bi-directional nonlinear learning

Objective. Accurate target localization is essential for effective radiation therapy, yet respiratory motion introduces considerable uncertainty. Ultrasound-based motion tracking offers a non-invasive solution, but the presence of similar anatomical structures and significant changes in appearance severely hinders robust and accurate tracking. Robust tracking is essential to ensure accuracy and improve treatment outcomes.Approach. We propose SiamAT, a tracking framework comprising a Siamese-like feature extraction network with a Swin-Transformer backbone, an attention-based target-aware module, and multi-task prediction heads. The target-aware module adaptively integrate template and search features to generate target-aware representations for precise localization.Main results. The proposed SiamAT is evaluated on a public dataset, i.e. MICCAI 2015 challenge on liver US tracking, and our clinical dataset provided by the Chinese People's Liberation Army General Hospital. Experimental results demonstrate that the method achieves accurate and robust tracking(0.60±0.30 mm and 0.47±0.31 mm tacking errors in two datasets) and outperforms existing methods. The proposed method runs at approximatively 36 fps on GPU.Significance. The proposed SiamAT provides clinicians with an effective tool for analyzing 2D motion in US images in clinical settings, assisting in the planning of optimal treatment strategies while reducing diagnostic and therapeutic risks.

Tracking and grasping a moving target is currently a challenging topic in the field of robotics. The current visual servo grasping method is still inadequate, as the real-time performance and robustness of target tracking both need to be improved. A target tracking method is proposed based on improved geometric particle filtering (IGPF). Following the geometric particle filtering (GPF) tracking framework, affine groups are proposed as state particles. Resampling is improved by incorporating an improved conventional Gaussian resampling algorithm. It addresses the problem of particle diversity loss and improves tracking performance. Additionally, the OTB2015 dataset and typical evaluation indicators in target tracking are adopted. Comparative experiments are performed using PF, GPF and the proposed IGPF algorithm. A dynamic target tracking and grasping method for the robot is proposed. It combines an improved Gaussian resampling particle filter algorithm based on affine groups and the positional visual servo control of the robot. Finally, the robot conducts simulation and experiments on capturing dynamic targets in the simulation environment and actual environment. It verifies the effectiveness of the method proposed in this paper.

This paper focuses on underwater target tracking based on a multi-static sonar network composed of passive sonobuoys and an active ping. In the multi-static sonar network, the location of the target can be estimated using TDOA (Time Difference of Arrival) measurements. However, since the sensor network may obtain insufficient and inaccurate TDOA measurements due to ambient noise and other harsh underwater conditions, target tracking performance can be significantly degraded. We propose a robust target tracking algorithm designed to operate in such a scenario. First, track management with track splitting is applied to reduce performance degradation caused by insufficient measurements. Second, a target location is estimated by a fusion of multiple TDOA measurements using a Gaussian Mixture Model (GMM). In addition, the target trajectory is refined by conducting a stack-based data association method based on multiple-frames measurements in order to more accurately estimate target trajectory. The effectiveness of the proposed method is verified through simulations.

The point detections obtained from radars or sonars in surveillance environments include clutter measurements, as well as target measurements. Target tracking with these data requires data association, which distinguishes the detections from targets and clutter. Various algorithms have been proposed for clutter measurement density estimation to achieve accurate and robust target tracking with the point detections. Among them, the spatial clutter measurement density estimator (SCMDE) computes the sparsity of clutter measurement, which is the reciprocal of the clutter measurement density. The SCMDE considers all adjacent measurements only as clutter, so the estimated clutter measurement density is biased for multi-target tracking applications, which may result in degraded target tracking performance. Through the study in this paper, a major source of tracking performance degradation with the existing SCMDE for multi-target tracking is analyzed, and the use of the clutter measurement probability is proposed as a remedy. It is also found that the expansion of the volume of the hyper-sphere for each sparsity order reduces the bias of clutter measurement density estimates. Based on the analysis, we propose a new adaptive clutter measurement density estimation method called SCMDE for multi-target tracking (MTT-SCMDE). The proposed method is applied to multi-target tracking, and the improvement of multi-target tracking performance is shown by a series of Monte Carlo simulation runs and a real radar data test. The clutter measurement density estimation performance and target tracking performance are also analyzed for various sparsity orders.

Video target tracking is a critical problem in the field of computer vision. Particle filters have been proven to be very useful in target tracking for nonlinear and non-Gaussian estimation problems. Although most existing algorithms are able to track targets well in controlled environments, it is often difficult to achieve automated and robust tracking of pedestrians in video sequences if there are various changes in target appearance or surrounding illumination. To surmount these difficulties, this paper presents multitarget tracking of pedestrians in video sequences based on particle filters. In order to improve the efficiency and accuracy of the detection, the algorithm firstly obtains target regions in training frames by combining the methods of background subtraction and Histogram of Oriented Gradient (HOG) and then establishes discriminative appearance model by generating patches and constructing codebooks using superpixel and Local Binary Pattern (LBP) features in those target regions. During the process of tracking, the algorithm uses the similarity between candidates and codebooks as observation likelihood function and processes severe occlusion condition to prevent drift and loss phenomenon caused by target occlusion. Experimental results demonstrate that our algorithm improves the tracking performance in complicated real scenarios.

Current single-object tracking algorithms depend on the information supplied by the template to identify and locate the object within the search area. However, environmental complexities and unknown factors can alter the object's state, causing mismatches in template information. The existing works using the template update mechanism (TUM) and multiple template feature fusion have the following problems: 1) TUM is affected by input superposition, making it hard to eliminate noise; 2) they suffer a temporal lag in their responsiveness to changes that occur in the object during the tracking process; 3) it is insufficient to rely solely on visual features within the search area of the current frame to improve the template; and 4) the prior knowledge regarding the input is not fully leveraged to learn the time-variant state of the object. We observe that in complex tracking scenarios, humans subconsciously analyze the evolutionary patterns of the object and its surroundings and integrate this information with the object's initial impression, thereby maintaining an awareness of the object's temporal state. Motivated by this, we propose a novel solution to the above problem, named MtvTrack, which can model the time-variant state of the object through the dynamic evolution pattern and static initial impression. Simultaneously, we propose a method for predicting the evolution pattern of scenes by utilizing past, present, and future (PPF) states. This approach effectively eliminates the information redundancy between consecutive frames and addresses the issue of delayed predictions of the target state in relation to changes within the search area. We establish a joint probability generative model and fully utilize prior knowledge to learn the time-variant state of the object. In addition, we develop a vector quantized-PPF (VQPPF) module for predicting the object's time-variant state. Experimental results on public benchmarks confirm the superior performance of our method. Source code is available at: https://github.com/long-wa/MtvTrack-main.

<div class="section abstract"><div class="htmlview paragraph">The camera-based advanced driver assistance systems (ADAS) like lane departure warning system (LDWS) and lane keeping assist (LKA) can make vehicles safer and driving easier. Lane detection is indispensable for these lane-based systems for achieving vehicle local localization and behavior prediction. Since the vision is vulnerable to the variable environment conditions such as bad weather, occlusions and illumination, the robustness is important. In this paper, a robust algorithm for detecting and tracking multiple lanes with arbitrary shape is proposed. We extend the previously lane detection and tracking process from the space domain to the temporal-spatial domain by using a more robust and general multi-lane model. First, new slice images containing temporal information are generated from image sequences. Instead of binarization process, we use a more general detector for extracting the lane marker candidates with prior knowledge to generate the binary slice image. Then, all the lane marker candidates are clustered into many lanes and as the initialization of the following particle filtering tracking process. We calculate the distance map in the binary slice image by using a modified distance algorithm and sample particles in the distance map. Finally, we can track the lane markers with particle filtering successfully. A range of experiment results indicate that the proposed algorithm can detect and track multiple lanes correctly and robustly. Comparing with other method, it has enough tolerance to variant illumination and occlusion.</div></div>

Shadows of moving objects often obstruct robust visual tracking. In this paper, we present a car tracker based on a hidden Markov model/Markov random field (HMM/MRF)-based segmentation method that is capable of classifying each small region of an image into three different categories: vehicles, shadows of vehicles, and background from a traffic-monitoring movie. The temporal continuity of the different categories for one small region location is modeled as a single HMM along the time axis, independently of the neighboring regions. In order to incorporate spatial-dependent information among neighboring regions into the tracking process, at the state-estimation stage, the output from the HMMs is regarded as an MRF and the maximum a posteriori criterion is employed in conjunction with the MRF for optimization. At each time step, the state estimation for the image is equivalent to the optimal configuration of the MRF generated through a stochastic relaxation process. Experimental results show that, using this method, foreground (vehicles) and nonforeground regions including the shadows of moving vehicles can be discriminated with high accuracy.

A registration-based tracking algorithm based on noise separation