Visual Tracking Algorithm Research Articles

Spiral visual and motional tracking

Constructing a visual appearance model is essential for visual tracking. However, relying only on the visual model during appearance changes is insufficient and may even interfere with achieving good results. Although several visual tracking algorithms emphasize motional tracking that estimates the motion state of the object center between consecutive frames, they suffer from accumulated error during runtime. As neither visual nor motional trackers are capable of performing well separately, several groups have recently proposed simultaneous visual and motional tracking algorithms. However, because tracking problems are often NP-hard, these algorithms cannot provide good solutions for the reason that they are driven top-down with low flexibility and often encounter drift problems. This paper proposes a spiral visual and motional tracking (SVMT) algorithm which, unlike existing algorithms, builds a strong tracker by cyclically combining weak trackers from both the visual and motional layers. In the spiral-like framework, an iteration model is used to search for the optimum until convergence, with the potential for achieving optimization. Three learned procedures including visual classification, motional estimation, and risk analysis are integrated into the generalized framework and implement corresponding modifications with regard to their performances. The experimental results demonstrate that SVMT performs well in terms of accuracy and robustness.

Occlusion-Aware Real-Time Object Tracking

The online learning methods are popular for visual tracking because of their robust performance for most video sequences. However, the drifting problem caused by noisy updates is still a challenge for most highly adaptive online classifiers. In visual tracking, target object appearance variation, such as deformation and long-term occlusion, easily causes noisy updates. To overcome this problem, a new real-time occlusion-aware visual tracking algorithm is introduced. First, we learn a novel two-stage classifier with circulant structure with kernel, named integrated circulant structure kernels (ICSK). The first stage is applied for transition estimation and the second is used for scale estimation. The circulant structure makes our algorithm realize fast learning and detection. Then, the ICSK is used to detect the target without occlusion and build a classifier pool to save these classifiers with noisy updates. When the target is in heavy occlusion or after long-term occlusion, we redetect it using an optimal classifier selected from the classifier-pool according to an entropy minimization criterion. Extensive experimental results on the full benchmark demonstrate our real-time algorithm achieves better performance than state-of-the-art methods.

Constrained Superpixel Tracking.

In this paper, we propose a constrained graph labeling algorithm for visual tracking where nodes denote superpixels and edges encode the underlying spatial, temporal, and appearance fitness constraints. First, the spatial smoothness constraint, based on a transductive learning method, is enforced to leverage the latent manifold structure in feature space by investigating unlabeled superpixels in the current frame. Second, the appearance fitness constraint, which measures the probability of a superpixel being contained in the target region, is developed to incrementally induce a long-term appearance model. Third, the temporal smoothness constraint is proposed to construct a short-term appearance model of the target, which handles the drastic appearance change between consecutive frames. All these three constraints are incorporated in the proposed graph labeling algorithm such that induction and transduction, short- and long-term appearance models are combined, respectively. The foreground regions inferred by the proposed graph labeling method are used to guide the tracking process. Tracking results, in turn, facilitate more accurate online update by filtering out potential contaminated training samples. Both quantitative and qualitative evaluations on challenging tracking data sets show that the proposed constrained tracking algorithm performs favorably against the state-of-the-art methods.

Exploiting superpixel and hybrid hash for kernel-based visual tracking

In recent years, significant advances in visual tracking have been made, and numerous outstanding algorithms have been proposed. However, the constraint between tracking accuracy and speed has not yet been comprehensively addressed. In this paper, to address the challenging aspects of visual tracking and, in particular, to achieve accurate real-time tracking, we propose a novel real-time kernel-based visual tracking algorithm based on superpixel clustering and hybrid hash analysis. By adopting superpixel clustering and segmentation, we reconstruct the appearance model of the target and its surrounding context in the initialization step. Via introducing the approach of overlap and intensity analysis, we divide the reconstructed model into several superpixel blocks. Based on the theory of circulant matrices and Fourier analysis, we build a Gaussian kernel correlation filter to roughly locate the position of each candidate block. To further improve the kernel correlation filter method, we compute each block's maximal response value in the confidence map and estimate each block's scale variation based on a peak value comparison. Additionally, we also propose a hybrid hash analysis strategy and integrate it with superpixel analysis for target blocks modification. By calculating a hybrid hash sequence based on L*A*B color and the discrete cosine transform, we conduct superpixel block modification to accurately locate the target and estimate the target's scale variation. Extensive experiments on visual tracking benchmark datasets show that our tracking algorithm outperforms the state-of-the-art algorithms and demonstrate its effectiveness and efficiency.

Compressive perceptual hashing tracking

This paper proposes a novel compressive sensing based perceptual hashing algorithm for visual tracking. A tracking object is represented by dimensionality reduced feature projected from perceptual hashing feature through a sparse measurement matrix. Besides, an updating weight map is assigned for each object and the weight map is updated according to the accumulation of foreground block and the distance between the foreground block and the center of the weight map. Based on above object representation and its weight map, our tracker searches the local region with the maximum similarity in coarse-to-fine way. In addition, we introduce a visual attention knowledge that the object, namely foreground, should be always located in the center of the weight map, to handle the model drift problem. Extensive experiments demonstrate that the proposed tracking method outperforms state-of-the-art methods in challenging scenarios and our tracker is especially insensitive to the location of the initial box.

Robust visual tracking based on scale invariance and deep learning

Visual tracking is a popular research area in computer vision, which is very difficult to actualize because of challenges such as changes in scale and illumination, rotation, fast motion, and occlusion. Consequently, the focus in this research area is to make tracking algorithms adapt to these changes, so as to implement stable and accurate visual tracking. This paper proposes a visual tracking algorithm that integrates the scale invariance of SURF feature with deep learning to enhance the tracking robustness when the size of the object to be tracked changes significantly. Particle filter is used for motion estimation. The confidence of each particle is computed via a deep neural network, and the result of particle filter is verified and corrected by mean shift because of its computational efficiency and insensitivity to external interference. Both qualitative and quantitative evaluations on challenging benchmark sequences demonstrate that the proposed tracking algorithm performs favorably against several state-of-the-art methods throughout the challenging factors in visual tracking, especially for scale variation.

Visual Object Tracking Based on Cross-Modality Gaussian-Bernoulli Deep Boltzmann Machines with RGB-D Sensors.

Visual object tracking technology is one of the key issues in computer vision. In this paper, we propose a visual object tracking algorithm based on cross-modality featuredeep learning using Gaussian-Bernoulli deep Boltzmann machines (DBM) with RGB-D sensors. First, a cross-modality featurelearning network based on aGaussian-Bernoulli DBM is constructed, which can extract cross-modality features of the samples in RGB-D video data. Second, the cross-modality features of the samples are input into the logistic regression classifier, andthe observation likelihood model is established according to the confidence score of the classifier. Finally, the object tracking results over RGB-D data are obtained using aBayesian maximum a posteriori (MAP) probability estimation algorithm. The experimental results show that the proposed method has strong robustness to abnormal changes (e.g., occlusion, rotation, illumination change, etc.). The algorithm can steadily track multiple targets and has higher accuracy.

Constructing a meta-tracker using Dropout to imitate the behavior of an arbitrary black-box tracker

Imitating the behaviors of an arbitrary visual tracking algorithm enables many higher level tasks such as tracker identification and efficient tracker-fusion. It is also useful for discovering the features essential in a black-box tracker or learning from several trackers to form a super-tracker. In this study, we propose a non-linear feature fusion framework, “MIMIC” that imitates many popular trackers by mixing a pool of heterogeneous features. The MIMIC framework consists of two subtasks, feature selection and feature weight tuning. These subtasks, however, tended to suffer from an overfitting problem when the number of videos available for training is limited. To address this issue, we incorporated Dropout algorithm into the training, which grants the trained MIMIC tracker a high degree of generalization. Extensive experiments testified the effectiveness of the proposed framework so that its applications would be promoted into different related tasks in visual tracking.

Performance Evaluation of Visual Tracking Algorithms on Video Sequences With Quality Degradation

Recently, there are lots of visual tracking algorithms proposed to improve the performance of object tracking in video sequences with various real conditions, such as severe occlusion, complicated background, fast motion, and so on. In real visual tracking systems, there are various quality degradation occurring during video acquisition, transmission, and processing. However, most existing studies focus on improving the accuracy of visual tracking while ignoring the performance of tracking algorithms on video sequences with certain quality degradation. In this paper, we investigate the performance evaluation of existing visual tracking algorithms on video sequences with quality degradation. A quality-degraded video database for visual tracking (QDVD-VT), including the reference video sequences and their corresponding distorted versions, is constructed as the benchmarking for robustness analysis of visual tracking algorithms. Based on the constructed QDVD-VT, we propose a method for robustness measurement of visual tracking (RMVT) algorithms by accuracy rate and performance stability. The performance of ten existing visual tracking algorithms is evaluated by the proposed RMVT based on the built QDVD-VT. We provide the detailed analysis and discussion on the robustness analysis of different visual tracking algorithms on video sequences with quality degradation from different distortion types. To visualize the robustness of visual tracking algorithms well, we design a robustness pentagon to show the accuracy rate and performance stability of visual tracking algorithms. Our initial investigation shows that it is still challenging for effective object tracking for existing visual tracking algorithms on video sequences with quality degradation. There is much room for the performance improvement of existing tracking algorithms on video sequences with quality degradation in real applications.

Visual Tracking Based on Complementary Learners with Distractor Handling

The representation of the object is an important factor in building a robust visual object tracking algorithm. To resolve this problem, complementary learners that use color histogram- and correlation filter-based representation to represent the target object can be used since they each have advantages that can be exploited to compensate the other’s drawback in visual tracking. Further, a tracking algorithm can fail because of the distractor, even when complementary learners have been implemented for the target object representation. In this study, we show that, in order to handle the distractor, first the distractor must be detected by learning the responses from the color-histogram- and correlation-filter-based representation. Then, to determine the target location, we can decide whether the responses from each representation should be merged or only the response from the correlation filter should be used. This decision depends on the result obtained from the distractor detection process. Experiments were performed on the widely used VOT2014 and VOT2015 benchmark datasets. It was verified that our proposed method performs favorably as compared with several state-of-the-art visual tracking algorithms.

基于图像签名算法的视觉目标跟踪算法

基于图像签名算法的视觉目标跟踪算法

基于可变形模型的目标跟踪算法

基于可变形模型的目标跟踪算法

基于深度学习的目标跟踪方法研究现状与展望

The inverse synthetic aperture lidar (ISAL) have attracted increasing attention for its merits including small visual tracking which is considered as one of the important research topics in the field of computer vision due to its key role in versatile applications, such as precision guidance, intelligent video surveillance, human-computer interaction, robot navigation and public safety. The basic idea for implementing visual tracking is composed of finding the target object in a video or sequence of images, then determining its exact position in the next successive frames and finally generating the corresponding trajectory of this object. Visual tracking, however, is still a challenging problem in practice while taking into account the abrupt appearance changes of the target objects induced by their non-rigid transformation, the sophisticated lighting variation, the obstruction by the block or similar objects in the background and the camera jitter. Motivated by the successful applications in target detection and recognition in recent years, plenty of deep learning models have been integrated in the visual tracking and better performance over traditional methods was achieved in a series of data evaluations, which opens a new door in the field of visual tracking. In this paper, the overview and progress on visual tracking were summarized. The current challenges and corresponding solving approaches in this field are introduced firstly and in particular, several novel and mainstream visual tracking algorithms based on the deep learning are specially described and analyzed in details, including their basic ideas, advantages and disadvantages and future prospect.

基于自适应卷积特征的目标跟踪算法

基于自适应卷积特征的目标跟踪算法

Visual Tracking via Sparse Representation with Reliable Structure Constraint

In this letter, we present a novel visual tracking algorithm based on sparse representation. In contrast to just use the target templates and the trivial templates to sparsely represent the target, we propose to further constrain the model with a set of discriminative weight maps. These weight maps contain the reliable structures of the target object. They help the model penalize the trivial template coefficients depending on the reliable structures of the target object. Then, the target object can be well represented by a sparse set of target templates together with a sparse set of target weight maps. We propose a unified objective function to integrate these two sparse representation problems together. This optimization problem can be well solved by the proposed iteration manner and a customized accelerated proximal gradient method. Furthermore, a novel weight map constructing method is proposed based on consistent motion property and forward–backward errors. Plenty of qualitative and quantitative evaluations demonstrate that our method performs favorably against the state-of-the-art methods in a wide range of tracking scenarios.

Visual tracking in HEVC compressed videos

We present a robust visual tracking algorithm to track a rigid object in high efficiency video coding (HEVC) compressed‐domain. Based on a Markov random field model, we estimate the target object state by using motion vector extracted from compressed bitstream. Especially, we employ residual discrete cosine transform (DCT) coefficient (RDC) in the bitstream as the shape information of the target object to prevent the drift problem. The measurements of motion vector and RDC are adaptively reflected through our potential functions for the target state. Our experiments on test sequences compressed by an HEVC encoder show that it is robust to track rigid objects. © 2016 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

Sparse representation combined with context information for visual tracking

In visual tracking, the main problem is to find candidates that are most likely to be the target in successive frames, so it is important to design a proper mechanism to evaluate this. In this paper, we propose a novel sparse representation based visual tracking algorithm, which well integrates the temporal and spatial context information of tracking objects into a unified framework. Specifically, we compute the similarity between the target and its candidates, which is acquired by fusing three aspects of the target's appearance variation with different weights. For the first part, we apply a patch based sparse representation to measure the similarities between the target in the first frame and candidates in current frame. Since the tracking result in the last frame provides the latest variation information of the target, we employ an image quality assessment method to obtain the similarity scores in the second part, and the spatial context information is also exploited. As the target appearance may suffer from radical changes along the video sequence, tracking that only uses the two parts mentioned above will suffer from serious drifting problems and easily cause incorrect tracking results. In order to ease this problem, we exploit the temporal context information by generating a group of history target templates adaptively according to previous tracking results, computing similarities between each candidate with them and the maximum will be used in the third part. Finally, we combine these parts to calculate the similarity scores and take those candidates with the highest score as the new targets in current frames. The extensive experiments on twelve challenging video sequences show that our algorithm can achieve performance competitive with state-of-the-art trackers.

Midlevel cues mean shift visual tracking algorithm based on target-background saliency confidence map

Because of the well-known merits of non-parametric estimation and fast mode matching, the mean shift tracking algorithm has been proposed with demonstrated success. But the traditional mean shift utilizes center-weighted color histogram as the reference model which is susceptible to interference of background pixels, and that can result in the compromised tracking robustness. In order to solve this problem, we propose a midlevel cues mean shift visual tracking algorithm based on target-background confidence map saliency-weighted model. A discriminative appearance model based on superpixels is introduced, thereby it can facilitate a tracker to distinguish between target and background by different weights. This improved mean shift tracker is formulated by computing a target-background saliency confidence map and mean shift iteration, and then we can obtain the position in next frame. Experimental results demonstrate that the improved mean shift tracker is able to handle occlusion and recover it from tracking drifts, furthermore, the improved algorithm facilitates foreground object segmentation during tracking.

Robust visual tracking based on generative and discriminative model collaboration

Effective object appearance model is one of the key issues for the success of visual tracking. Since the appearance of a target and the environment changes dynamically, the majority of existed visual tracking algorithms tend to drift away from targets. To address this issue, we propose a robust tracking algorithm by integrating the generative and discriminative model. The object appearance model is made up of generative target model and a discriminative classifier. For the generative target model, we adopt the weighted structural local sparse appearance model combining patch based gray value and Histogram of Oriented Gradients feature as the patch dictionary. By sampling positives and negatives, alignment-pooling features are obtained based on the patch dictionary through local sparse coding, then we use support vector machine to train the discriminative classifier. The proposed method is embedded into a Bayesian inference framework for visual tracking. A combined matching method is adopted to improve the proposal distribution of the particle filter. Moreover, in order to adapt the situation change, the patch dictionary and discriminative classifier are updated by incremental learning every five frames. Experimental results on some publicly available benchmarks of video sequences demonstrate the accuracy and effectiveness of our tracker.

Non-linear eigenspace visual object tracking

This paper presents a visual object tracking algorithm using an eigenspace representation. Previous approaches to eigenspace methods for object tracking use vectorized image regions as observations. Here, feature vectors associated to pixels of the target template are considered to be individual observations of the target object. The collection of observations is learned using non-linear subspace projection to arrive at an eigenspace representation. This representation allows tracking pixel-wise but the pixels are tied together using subspace representation which provides a robust and compact representation of the object. Localization and segmentation are carried out by deriving a similarity function in the eigenspace representation. A gradient descent and mean-shift based techniques are derived to maximize the similarity function with respect to the transformation parameters. The de-noising and clustering capabilities of the eigenspace representation lead to a localization procedure that is robust to noise and partial occlusion. For fast moving objects and to recover from total occlusion, a probabilistic search strategy, based on particle filter, is also developed. A unique feature of our approach is that it permits segmentation in addition to localization when multiple templates of the target are given. The performance of the algorithm is tested on real world tracking examples.