改进的粒子滤波器目标跟踪方法

针对现有的粒子滤波跟踪方法存在的不足,提出了一种改进的粒子滤波器方法用于运动目标跟踪.将颜色直方图和边缘直方图结合起来建立目标的参考模型,有效地克服了使用单一特征建模的缺点,提高了跟踪的准确性.分别计算目标颜色模型和目标边缘模型与粒子的欧几里德距离,使用这2个距离作为粒子权值计算的重要依据.实验结果表明该算法具有较高的实时性、准确性和鲁棒性.     

多相关滤波自适应融合的鲁棒目标跟踪

目的 由于目标在复杂场景中可能会发生姿态变化、物体遮挡、背景干扰等情况,目标跟踪仍然是一个具有挑战性的课题。目前判别性相关滤波方法在目标跟踪问题上获得了成功而又广泛的应用。标准的相关滤波方法基于循环偏移得到大量训练样本,并利用快速傅里叶变换加速求解滤波器,使其具有很好的实时性和鲁棒性,但边界偏移带来的消极的训练样本降低了跟踪效果。空间正则化的相关滤波跟踪方法引入空间权重函数,增强目标区域的滤波器作用,在增大了目标搜索区域的同时,也增加了计算时间,而且对于目标形变不规则,背景相似的情景也会增强背景滤波器,从而导致跟踪失败。为此,基于以上问题,提出一种自适应融合多种相关滤波器的方法。方法 利用交替方向乘子法将无约束的相关滤波问题转化为有约束问题的两个子问题,在子问题中分别采用不同的相关滤波方法进行求解。首先用标准的相关滤波方法进行目标粗定位,进而用空间正则化的相关滤波跟踪方法进行再定位,实现了目标位置和滤波模板的微调,提高了跟踪效果。结果 本文算法和目前主流的一些跟踪方法在OTB-2015数据集中100个视频上,以中心坐标误差和目标框的重叠率为评判标准进行了对比实验,本文算法能较好地处理多尺度变化、姿态变化、背景干扰等问题,在CarScale、Freeman4、Girl等视频上都表现出了最好的跟踪结果;本文算法在100个视频上的平均中心坐标误差为28.55像素,平均目标框重叠率为61%,和使用人工特征的方法相比,均高于其他算法,与使用深度特征的相关滤波方法相比,平均中心坐标误差高了6像素,但平均目标框的重叠率高了4%。结论 大量的实验结果表明,在目标发生姿态变化、尺度变化等外观变化时,本文算法均具有较好的准确性和鲁棒性。     

Sparse Bayesian learning for efficient visual tracking

This paper extends the use of statistical learning algorithms for object localization. It has been shown that object recognizers using kernel-SVMs can be elegantly adapted to localization by means of spatial perturbation of the SVM. While this SVM applies to each frame of a video independently of other frames, the benefits of temporal fusion of data are well-known. This is addressed here by using a fully probabilistic relevance vector machine (RVM) to generate observations with Gaussian distributions that can be fused over time. Rather than adapting a recognizer, we build a displacement expert which directly estimates displacement from the target region. An object detector is used in tandem, for object verification, providing the capability for automatic initialization and recovery. This approach is demonstrated in real-time tracking systems where the sparsity of the RVM means that only a fraction of CPU time is required to track at frame rate. An experimental evaluation compares this approach to the state of the art showing it to be a viable method for long-term region tracking.     

Particle filter re-detection for visual tracking via correlation filters

Most of the correlation filter based tracking algorithms can achieve good performance and maintain fast computational speed. However, in some complicated tracking scenes, there is a fatal defect that causes the object to be located inaccurately, which is the trackers excessively dependent on the maximum response value to determine the object location. In order to address this problem, we propose a particle filter redetection based tracking approach for accurate object localization. During the tracking process, the kernelized correlation filter (KCF) based tracker can locate the object by relying on the maximum response value of the response map; when the response map becomes ambiguous, the tracking result becomes unreliable correspondingly. Our redetection model can provide abundant object candidates by particle resampling strategy to detect the object accordingly. Additionally, for the target scale variation problem, we give a new object scale evaluation mechanism, which merely considers the differences between the maximum response values in consecutive frames to determine the scale change of the object target. Extensive experiments on OTB2013 and OTB2015 datasets demonstrate that the proposed tracker performs favorably in relation to the state-of-the-art methods.

     

基于改进Bhattacharyya系数的粒子滤波视觉跟踪算法

基于颜色直方图的粒子滤波跟踪通常采用Bhattacharyya系数(B氏系数)衡量目标与候选区域特征模型之间的相似性.分析说明目标内部区域的B氏系数存在大量的峰值,使得粒子滤波跟踪仅能适应目标收缩,无法适应目标的膨胀.为此,提出了一种改进的B氏系数,从理论上分析说明了该系数具有单峰特性,基于该系数的粒子滤波跟踪能同时适应目标收缩和膨胀.分析和实验结果均表明,基于本文提出的改进B氏系数的粒子滤波跟踪对目标快速膨胀和收缩等形变具有较好的鲁棒性和准确性.     

基于空间直方图的粒子滤波目标跟踪算法

空间直方图融合了目标的颜色信息和颜色的空间分布信息,比传统的颜色直方图更具有目标鉴别能力。在基于粒子滤波算法的目标跟踪系统框架中,采用简单的随机漂移模型表示系统状态模型,通过空间直方图的相似度定义来建立系统观测概率模型,提出一种基于空间直方图的粒子滤波目标跟踪算法。实验结果表明,相比传统的基于颜色直方图的粒子滤波算法,提出的算法具有更好的鲁棒性。     

Dynamic appearance model for particle filter based visual tracking

This paper made major research on the target representation problem, which plays a significant role in visual tracking, but has received little attention in most researches. In order to fulfill the requirements of tracking robustness and effectiveness in practical conditions, a dynamic appearance model is constructed. Due to particle filter's excellent characteristics, it is employed in this paper not only to estimate target's state, but also to construct the dynamic observation model integrated by multiple cues. In the proposed method, a dynamic multi-cue integration model is constructed for particle filter framework. And a systematic study is done on evaluating cue's weight. Specially, a particle filter based weight tracker is designed to update multi-cue's integrating manner online, so as to adapt the observation model to target's appearance changes. In such a way, a double-particle-filter based tracking framework is formed, and it is field tested on a variety of videos in different tracking conditions. In the experiments and comparisons, the applicable conditions of the proposed dynamic model are discussed, and its robustness and effectiveness are demonstrated.     

Adaptive weighting of local classifiers by particle filters for robust tracking

This paper presents an adaptive weighting method for combining local classifiers using a particle filter. Although the effectiveness of weighting methods based on combinations of local classifiers (features) has been reported recently, such methods fail in cases where there is partial occlusion or when shadows appear due to changes in the illumination direction since fixed weights are used for combining the local classifiers. In order to achieve the desired robustness, the weights should be changed adaptively. For this purpose, we use a particle filter, where each particle is assigned to the weight set for combining local classifiers. By estimating the posterior probability in weight space by using a particle filter, the effective weights for current time-step are obtained, and as a result the proposed method can account for dynamic occlusion. As a means of a demonstration, our approach is applied to the face tracking problem. The adaptability and the robustness of the method with respect to partial occlusion are evaluated using test sequences in which the occluded areas are changed dynamically. The weights of the occluded regions decrease automatically without the need for explicit knowledge about the occurrence of occlusion, which makes it possible to track the face under conditions of dynamic occlusion.     

An efficient visual tracking method for multiple moving targets

An efficient algorithm of the edge detection according to integrating the edge gradient with the average filter is proposed, which can significantly reduce sensitivity of the background subtraction method to noise and illumination. Taking into account the features of the target such as color, size, etc., a new modified Nearest Neighbor （NN） algorithm for data association using the target features is designed. A designed Interacting Multiple Model （IMM） filter is utilized to track the maneuvering target motion, i.e. the feature point （called the centroid of the target） motion of the target. The algorithms are validated via an example with natural video sequences. The results show the algorithms are performances and validity for visual tracking. In complex environment, the algorithm can still work well.     

Multiple model target tracking with variable rate particle filters

Fixed rate state space models are the conventional models used to track the maneuvering objects. In contrast to fixed rate models, recently introduced variable rate particle filter (VRPF) is capable of tracking the target with a small number of states by imposing a Gamma distribution on the state arrival times while the object trajectory is approached by a single dynamic motion model. Using a single dynamic motion model limits the capability of estimating the characteristics of maneuvering and smooth regions of the trajectory. To overcome this weakness we introduce an adaptive tracking method which incorporates multiple model approach with the variable rate model structure. The proposed model referred to as multiple model variable rate particle filter (MM-VRPF) adaptively locates frequent state points to the maneuvering regions resulting in a much more accurate tracking while preserving the parsimonious representation for the smooth regions of the trajectory. This is achieved by including a mode variable into the conventional variable rate state vector that enables us to define different sojourn and motion parameters for each motion mode using the multiple model structure. Simulation results show that the proposed algorithm outperforms the conventional variable rate particle filter, fixed rate multiple model particle filter and interacting multiple model.     

Real-time visual tracking using complementary kernel support correlation filters

Despite demonstrated success of SVM based trackers,their performance remains a boosting room if carefully considering the following factors:first,the tradeoff between sampling and budgeting samples affects tracking accuracy and efficiency much;second,how to effectively fuse different types of features to learn a robust target representation plays a key role in tracking accuracy.In this paper,we propose a novel SVM based tracking method that handles the first factor with the help of the circulant structures of the samples and the second one by a multi-kernel learning mechanism.Specifically,we formulate an SVM classification model for visual tracking that incorporates two types of kernels whose matrices are circulant,fully taking advantage of the complementary traits of the color and HOG features to learn a robust target representation.Moreover,it is fortunate that the SVM model has a closed-form solution in terms of both the classifier weights and the kernel weights,and both can be efficiently computed via fast Fourier transforms(FFTs).Extensive evaluations on OTB100 and VOT2016 visual tracking benchmarks demonstrate that the proposed method achieves a favorable performance against various state-of-the-art trackers with a speed of 50 fps on a single CPU.     

Distractor-aware visual tracking using hierarchical correlation filters adaptive selection

Applied Intelligence - In recent years, the ensembled trackers composed of multi-level features from the pre-trained Convolutional Neural Network (CNN) have achieved top performance in visual...     

A multiple feature fused model for visual object tracking via correlation filters

Multimedia Tools and Applications - Common tracking algorithms only use a single feature to describe the target appearance, which makes the appearance model easily disturbed by noise. Furthermore,...     

Multi-dimensional visual tracking using scatter search particle filter

Multi-dimensional visual tracking (MVT) problems include visual tracking tasks where the system state is defined by a high number of variables corresponding to multiple model components and/or multiple targets. A MVT problem can be modeled as a dynamic optimization problem. In this context, we propose an algorithm which hybridizes particle filters (PF) and the scatter search (SS) metaheuristic, called scatter search particle filter (SSPF), where the optimization strategies from SS are embedded into the PF framework. Scatter search is a population-based metaheuristic successfully applied to several complex combinatorial optimization problems. The most representative optimization strategies from SS are both solution combination and solution improvement. Combination stage enables the solutions to share information about the problem to produce better solutions. Improvement stage makes also possible to obtain better solutions by exploring the neighborhood of a given solution. In this paper, we have described and evaluated the performance of the scatter search particle filter (SSPF) in MVT problems. Specifically, we have compared the performance of several state-of-the-art PF-based algorithms with SSPF algorithm in different instances of 2D articulated object tracking problem and 2D multiple object tracking. Some of these instances are from the CVBase’06 standard database. Experimental results show an important performance gain and better tracking accuracy in favour of our approach.     

利用视觉显著性和粒子滤波的运动目标跟踪

针对运动目标跟踪问题,提出一种利用视觉显著性和粒子滤波的目标跟踪算法.借鉴人类视觉注意机制的研究成果,根据目标的颜色、亮度和运动等特征形成目标的视觉显著性特征,与目标的颜色分布模型一起作为目标的特征表示模型,利用粒子滤波进行目标跟踪.该算法能够克服利用单一颜色特征所带来的跟踪不稳定问题,并能有效解决由于目标形变、光照变化以及目标和背景颜色分布相似而产生的跟踪困难问题,具有较强的鲁棒性.在多个视频序列中进行实验,并给出相应的实验结果和分析.实验结果表明,该算法用于实现运动目标跟踪是正确有效的.     

Multi-robot cooperative spherical-object tracking in 3D space based on particle filters

This article presents a cooperative approach for tracking a moving spherical object in 3D space by a team of mobile robots equipped with sensors, in a highly dynamic environment. The tracker’s core is a particle filter, modified to handle, within a single unified framework, the problem of complete or partial occlusion for some of the involved mobile sensors, as well as inconsistent estimates in the global frame among sensors, due to observation errors and/or self-localization uncertainty. We present results supporting our approach by applying it to a team of real soccer robots tracking a soccer ball, including comparison with ground truth.     

Real-time visual tracking via multi-cue based adaptive particle filter framework

Visual tracking using particle filter has been extensively investigated due to its myriad of application in the field of computer vision. However, particle filter framework performance is heavily impaired due to its inherent problems namely, particle degeneracy and impoverishment. In addition, most of the tracking methods using single cue are greatly affected by dynamic environmental challenges. To address these issues, we propose an adaptive multi-cue particle filter based real-time visual tracking framework. Three complementary cues namely, color histogram, LBP and pyramid of histogram of gradient have been exploited for object’s appearance model. These cues are integrated using the proposed adaptive fusion model for the automatic boosting of important particles and suppression of unimportant particles. Resampling method using butterfly search optimization relocate low performing particles to high likelihood area. Proposed outlier detection mechanism not only helps in detecting low performing particles but also aids in updating of the reference dictionary. Online estimation of cue reliability along with its multi-cue fusion leads to quick adaptation of the proposed tracker. On average of the outcome, our tracker achieves average center location error of 6.89 (in pixels) and average F-measure of 0.786 when evaluated on OTB-100 and VOT dataset against 13 others state-of-the-art.

     

Consistency checks for particle filters

An "inconsistent" particle filter produces - in a statistical sense - larger estimation errors than predicted by the model on which the filter is based. Two test variables are introduced that allow the detection of inconsistent behavior. The statistical properties of the variables are analyzed. Experiments confirm their suitability for inconsistency detection.     

基于稀疏卷积特征和相关滤波的实时视觉跟踪算法

为提高分层卷积相关滤波视觉跟踪算法的实时性能,提出一种稀疏卷积特征的实时目标跟踪算法。首先,在分析不同层卷积特征的基础上,采用等间隔采样的方式提取每个卷积层的稀疏卷积特征;然后,对每个卷积层特征的相关滤波响应值进行加权组合,得到目标预测的位置;最后,采用稀疏的模型更新策略进一步提高算法的运行速度。在OTB-2015新增的50组数据上对所提算法进行测试,实验结果表明,该算法的平均距离精度为82.2%,比原分层卷积特征跟踪算法提高了5.25个百分点,对目标姿态以及遮挡等变化具有较好的鲁棒性。该算法的平均跟踪速度为32.6帧/s,是原分层卷积特征跟踪算法的近3倍,能达到实时跟踪的效果。