Effects of Improper Ground Truth on Target Tracking Performance Evaluation in Benchmark

In this paper, we present a framework for tracking a moving target in urban environments using UAVs in cooperation with UGVs. The framework takes into account occlusions between the sensor and the target. The target state is modeled using the dynamic occupancy grid and the target motion model is built using a second-order Markov chain. Based on the target occupancy grid, we design the path planning algorithm to maneuver the UAV and the UGV to configurations where they can detect the target with high probability. Simulation results show the framework is successful in solving the target tracking problem in urban environments. We also build an indoor hardware platform and successfully implement preliminary tracking algorithms on the platform. Results show the platform can be further used to test more advanced tracking algorithms like the proposed tracking framework.

The Kalman filter has long been regarded as the optimal solution to many applications in computer vision for example the tracking objects, prediction and correction tasks. Its use in the analysis of visual motion has been documented frequently, we can use in computer vision and open cv in different applications in reality for example robotics, military image and video, medical applications, security in public and privacy society, etc. In this paper, we investigate the implementation of a Matlab code for a Kalman Filter using three algorithm for tracking and detection objects in video sequences (block-matching (Motion Estimation) and Camshift Meanshift (localization, detection and tracking object)). The Kalman filter is presented in three steps: prediction, estimation (correction) and update. The first step is a prediction for the parameters of the tracking and detection objects. The second step is a correction and estimation of the prediction parameters. The important application in Kalman filter is the localization and tracking mono-objects and multi-objects are given in results. This works presents the extension of an integrated modeling and simulation tool for the tracking and detection objects in computer vision described at different models of algorithms in implementation systems.

In this paper an advanced anti-jam indoor adaptive GNSS signal acquisition and tracking algorithm is considered. Initially, we were able to determine that double-dwell structure (DDS) reduces the processing time penalty caused by false alarm. Nevertheless, the double-dwell structure is still vulnerable to interference and jamming. In order to determine a suitable advanced anti-jam indoor adaptive GNSS signal acquisition and tracking algorithm for DDS we first perform the Bayesian parameter estimation; i.e., we analytically compute the posterior Bayes probability density function (pdf) and cumulative distribution function (cdf) by applying the Bayes theorem in three steps. First, we compute the complex signal distribution and complex matrix variate signal distribution. This is an original new result never published before. Second, we provide an introduction of the complex Bessel or parabolic function interference distribution due to earlier assumptions of the DDS interference distribution models. We further explain that the process that is required to produce Complex Matrix Variate Bessel Interference Distribution or Complex Matrix Variate Parabolic Function Interference Distribution is an extremely daunting task let alone the Complex Matrix Variate DDS Bayesian posterior density it would seem at the moment nearly an impossible task because it may require the computation of functions such as Kampe de Feriet function and Jack functions of matrix arguments. Third, instead of performing the second step, we relax the assumption of interference to normal distribution. In both the scalar case and complex matrix variate cases we observe that the complex matrix variate Bayesian posterior pdf or cdf is invariant of the observation data or is identical to the prior complex matrix variate signal distribution model. This is an original new and very powerful result never published before. Why this result is so powerful is because up until now we never had a complete theoretical validation of our GNSS receiver design based on either autocorrelation or cross-correlation properties since, the complex matrix variate Bayesian posterior pdf or cdf is invariant of the observation data or is identical to the prior complex matrix variate signal distribution model. Future simulation results will illustrate that an anti-jam indoor adaptive DDS stable detection structure reduces by half the average acquisition time and significantly outperform its predecessor against interference and jamming. Moreover, future simulation results will show further advancements of the Giftet Inc. MATLAB library capability to perform advanced numerical computations based on closed form expressions of the generalized Bessel function distributions and generalized parabolic cylinder function distribution models via Kampe de Feriet function and Jack functions.

Target tracking is a hot topic in the field of computer vision. It uses the context information of video or image sequence to model the appearance and motion information of the target, so as to predict the motion state of the target and calibrate the position of the target. From the perspective of building models, target tracking algorithms can be divided into generative models and discriminant models; From the number of tracking targets, it can be divided into single target tracking and multi-target tracking. Target tracking integrates the theories and algorithms of image processing, machine learning, optimization and other fields. It is the premise and foundation to complete higher-level image understanding.The basic task of target tracking is to give the initial position of the target in a video sequence and continuously track and locate the target in each subsequent frame. In this process, a priori conditions such as the color, shape and size of the target will not be provided, that is, the tracking algorithm can track only by learning the target in the first frame. The traditional solution is to calculate the project similarity offline and save it in the system for algorithm call, but it can not make full use of the latest scoring data to reflect the change of user interest. To solve the above problems, a target tracking item similarity incremental update mechanism suitable for online applications is proposed, so that the recommendation system can update the similarity data between the corresponding item and other items in real time after the current user submits the item score, so that the recommendation system can recommend based on the latest item similarity data, To adapt to changes in user interests.

Single-object tracking, also known as visual tracking, on the drone platform attracts much attention recently with various applications in computer vision, such as filming and surveillance. However, the lack of commonly accepted annotated datasets and standard evaluation platform prevent the developments of algorithms. To address this issue, the Vision Meets Drone Single-Object Tracking (VisDrone-SOT2018) Challenge workshop was organized in conjunction with the 15th European Conference on Computer Vision (ECCV 2018) to track and advance the technologies in such field. Specifically, we collect a dataset, including 132 video sequences divided into three non-overlapping sets, i.e., training (86 sequences with 69, 941 frames), validation (11 sequences with 7, 046 frames), and testing (35 sequences with 29, 367 frames) sets. We provide fully annotated bounding boxes of the targets as well as several useful attributes, e.g., occlusion, background clutter, and camera motion. The tracking targets in these sequences include pedestrians, cars, buses, and animals. The dataset is extremely challenging due to various factors, such as occlusion, large scale, pose variation, and fast motion. We present the evaluation protocol of the VisDrone-SOT2018 challenge and the results of a comparison of 22 trackers on the benchmark dataset, which are publicly available on the challenge website: http://www.aiskyeye.com/. We hope this challenge largely boosts the research and development in single object tracking on drone platforms.

InceptB: A CNN Based Classification Approach for Recognizing Traditional Bengali Games

Deformable models are a useful tool in computer vision and computer graphics. A deformable model is a curve (in two dimensions) or a surface (in three dimensions), whose shape, position, and orientation are controlled through a set of parameters. Deformable models can represent manufactured objects, human faces and skeletons, and even bodies of fluid. In computer graphics, we use deformable models for animations and simulations, whereas in computer vision applications, such as tracking and fitting, deformable models help to restrict the family of possible solutions. We introduce the use of a directed acyclic graph (DAG) to describe the position and Jacobian of each point on the surface of deformable models. This data structure, combined with a topological description of the points, is simple, powerful, and extremely useful for both computer vision and computer graphics applications. We show a computer vision application, 3D deformable face tracking, and a computer graphics application, cyberglove data visualization and calibration.

Segmentation of independent motions from video sequences is a challenging problem that can be a prelude to many further applications in computer vision. In this paper, we present an accurate and efficient approach for automatic segmentation of all the independently moving objects in the scene. The system begins with an initialization module which provides initial partition of the scene, computes 2D motions, and includes some necessary preparing work. Then, we propose a novel object classification method by analyzing and clustering motions. To achieve the best robustness, and minimize the total computation load, we choose to work on multi key frames simultaneously to obtain global optimal classification. Our approach achieves accurate object classification and avoids the uncertainty in detection of moving objects. We demonstrate high stability, accuracy and performance of our algorithm with a set of experiments on real video sequences.

Intrinsic image decomposition is an important topic in computer vision and computer graphics applications. However, this is a challenging problem by adopting the information of a single image. Therefore, additional priors or supplementary information such as multiply images or user interactions are necessary to address this problem. In this paper, we propose a novel scheme to use multiple images for intrinsic image decomposition, based on a similar strategy in robust principal component analysis (RPCA). RPCA utilizes the fact that the reflectance layer is common in multiple lights of a scene, and attempts to decompose the data matrix constructed from input images into a low-rank matrix and a sparse matrix. This is possible if the sparse matrix is sufficiently sparse, which is often not the case in computer vision applications. Moreover, the weighting parameter between the low-rank and sparse matrices greatly affects the accuracy of the results, and tuning this parameter can be tricky. This paper proposes a rank-constrained PCA algorithm (RCPCA) for solving background recovering problems. Fixing the rank of the low-rank matrix to be 1 allows RCPCA to better recover the low-rank matrix from the data matrix. Comprehensive tests show that RCPCA produces more stable and accurate results than RP-CA.

This paper presents a novel framework for importance-aware adaptive data transmission, designed specifically for real-time computer vision (CV) applications where taskspecific fidelity is critical. An importance-weighted mean square error (IMSE) metric is introduced, assigning data importance based on bit positions within pixels and semantic relevance within visual segments, thus providing a task-oriented measure of reconstruction quality. To minimize IMSE under the total power constraint, a data-importance-aware waterfilling approach is proposed to optimally allocate transmission power according to data importance and channel conditions. Simulation results demonstrate that the proposed approach significantly outperforms margin-adaptive waterfilling and equal power allocation strategies, achieving more than 7 dB and 10 dB gains in normalized IMSE at high SNRs (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$&gt;10 ~\text{dB}$</tex>), respectively. These results highlight the potential of the proposed framework to enhance data efficiency and robustness in real-time CV applications, especially in bandwidth-limited and resource-constrained environments.

english

991013340459703412 HKUST Electronic Theses Open-set recognition and its applications in computer vision by Jun Cen thesis 2024 1 online resource (xiv, 92 pages) : illustrations (chiefly color) Current deep learning models are trained to fit the training set distribution. Despite…Read more ›

The aims of this paper is to propose a novel method for solving problem face detection and tracking system based in four algorithms: block-matching (BMA), Meanshift, Camshift and Kalman filter. Firstly, BMA is applied to the different sequential test as a preprocessing stage to detect faces. Then, the face tracking system are with three modules: Meanshift, Camshift and Kalman filter. This scheme gives a better face detection and tracking. Our work increases the performance and other criteria values in the embedded system. The human is more and more interested in producing intelligence which is one of the most impressive features of natures. Researchers are trying to make intelligent machine that have various capabilities. Building a machine or robot is probably one of the most challenging problems which humans are trying to solve. Recently, many projects have started with the purpose of learning machines to track some particular objects. One of the most challenging applications in computer vision is tracking objects efficiently in video sequence. Though progress has been accomplished, the best algorithms are far from reaching the speed and the performance of system. Object or multi-object tracking (face, human, car, etc.) is a fundamental problem that merits particular attention, since it is the key to solve a number of computer vision applications.

Transformer is a kind of deep neural network that relies on the technique of self-attention and used initially in the field of natural language processing. Scientists use transformer for computer vision (CV) applications because of its good data representation capabilities. Transformer-based models yield similar performance or surpass other network architectures, including convolutional and recurrent neural networks, in a variety of visual benchmarks. In this work, we investigate the methods for video anomaly detection (VAD) using vision transformer models in the recent literature. The main topics we explore comprise vision transformers used in CV applications with a special focus on VAD methods leveraging transformer architecture. We also briefly present anomaly detection methods based on transformers. Additionally, we address the advantages, challenges and current limitations of the transformer architecture as well as potential solutions to address the technical challenges. In the concluding section of this study, we offer avenues for further investigation concerning the use of vision transformers in VAD tasks.

Object tracking is one of the most important topics in computer vision. While the state-of-the-art tracking algorithms achieved great success, there are still some challenging problems to be solved. Firstly, it remains a tough task to develop a tracking algorithm with both accuracy and efficiency. Secondly, the ground truth is often given by a rectangular bounding box, which contains not only the target, but also the background pixels. The background pixels in the initial bounding box will mislead the appearance model of the target. Thirdly, most features for representing the target only use the gray scale information, and are not robust to pedestrians. The third problem is especially serious when the targets are pedestrians, which have colorful cloths on them, and change their pose and shape while walking. In this paper, we propose a novel saliency-based H-S histogram algorithm. Saliency detection can efficiently delete most of the background pixels, which makes the appearance model more accurate. H-S histogram is a statistic-based feature, which is more robust to shape deformation, and the color information is considered in the H-S channels. Experiments on Caltech pedestrian database show the proposed method can handle some hard cases, and achieves a higher success rate.