Target Scale Variation Research Articles

An Improved Kernelized Correlation Filter Based Visual Tracking Method

Correlation filter based trackers have received great attention in the field of visual target tracking, which have shown impressive advantages in terms of accuracy, robustness, and speed. However, there are still some challenges that exist in the correlation filter based methods, such as target scale variation and occlusion. To deal with these problems, an improved kernelized correlation filter (KCF) tracker is proposed, by employing the GM(1,1) grey model, the interval template matching method, and multiblock scheme. In addition, a strict template update strategy is presented in the proposed method to accommodate the appearance change and avoid template corruption. Finally, some experiments are conducted. The proposed method is compared with the top state-of-the-art trackers, and all the tracking algorithms are evaluated on the object tracking benchmark. The experimental results demonstrate obvious improvements of the proposed KCF-based visual tracking method.

Real-time visual tracking with ELM augmented adaptive correlation filter

Accuracy, robustness and efficiency are three general requirements for the online visual tracking. Although the correlation based trackers achieved competitive results on both accuracy and efficiency, they still suffer from time-consuming scale searching and are prone to drifting. In this paper, an Extreme Learning Machine (ELM) augmented adaptive correlation filter is proposed to address these issues. We enhance the correlation based tracker in four aspects. Firstly, a keypoints tracking based scale estimation method is proposed to handle the target scale variation during tracking. Secondly, the powerful and efficient features including intensity, Fast Histogram of Oriented Gradient (FHOG) and Discriminative color Descriptors (DD) are integrated to further boost the tracking performance. Thirdly, dynamic updating mechanism is introduced to better adapt the correlation filter to the appearance changes. Moreover, to handle the drift brought by severe occlusion, we train an online ELM based classifier to re-initialize object in case of tracking failure. Extensive experimental results on the Object Tracking Benchmark (OTB) dataset show that our proposed tracker performs favorably against state-of-the-art methods in terms of accuracy, efficiency, and robustness.

Color Feature-Based Object Tracking through Particle Swarm Optimization with Improved Inertia Weight.

This paper presents a particle swarm tracking algorithm with improved inertia weight based on color features. The weighted color histogram is used as the target feature to reduce the contribution of target edge pixels in the target feature, which makes the algorithm insensitive to the target non-rigid deformation, scale variation, and rotation. Meanwhile, the influence of partial obstruction on the description of target features is reduced. The particle swarm optimization algorithm can complete the multi-peak search, which can cope well with the object occlusion tracking problem. This means that the target is located precisely where the similarity function appears multi-peak. When the particle swarm optimization algorithm is applied to the object tracking, the inertia weight adjustment mechanism has some limitations. This paper presents an improved method. The concept of particle maturity is introduced to improve the inertia weight adjustment mechanism, which could adjust the inertia weight in time according to the different states of each particle in each generation. Experimental results show that our algorithm achieves state-of-the-art performance in a wide range of scenarios.

Online Learning of Discriminative Correlation Filter Bank for Visual Tracking

Accurate visual tracking is a challenging research topic in the field of computer vision. The challenge emanates from various issues, such as target deformation, background clutter, scale variations, and occlusion. In this setting, discriminative correlation filter (DCF)-based trackers have demonstrated excellent performance in terms of speed. However, existing correlation filter-based trackers cannot handle major changes in appearance due to severe occlusions, which eventually result in the development of a bounding box for target drift tracking. In this study, we use a set of DCFs called discriminative correlation filter bank (DCFB) for visual tracking to address the key causes of object occlusion and drift in a tracking-by-detection framework. In this work, we treat thxe current location of the target frame as the center, extract several samples around the target, and perform online learning of DCFB. The sliding window then extracts numerous samples within a large radius of the area where the object in the next frame is previously located. These samples are used for the DCFB to perform correlation operation in the Fourier domain to estimate the location of the new object; the coordinates of the largest correlation scores indicate the position of the new target. The DCFB is updated according to the location of the new target. Experimental results on the quantitative and qualitative evaluations on the challenging benchmark sequences show that the proposed framework improves tracking performance compared with several state-of-the-art trackers.

Robust Object Tracking via Large Margin and Scale-Adaptive Correlation Filter

Due to the impressive performance and computational efficiency of correlation filter (CF)-based object tracking methods, CF trackers have gained lots of popularity in recent years. However, target drift and tracking failure caused by background clutter and target appearance change (resulting from scale variation and deformation and so on) are still challenging tasks. To overcome these challenges, we propose a new tracking method within the CF framework in this paper. First, we learn a large margin CF by exploiting discriminative background patches. Contrary to conventional CF trackers that aim to maximize target response, we model a tracker that maximizes the margin between the target and surrounding background by exploiting background information effectively. To remedy the deficiency in handling target scale variation of CF-based trackers, we propose to train a CF by multi-level scale supervision, which aims to make CF sensitive to the target scale variation. Then, we integrate the two individual modules into one framework to simplify our tracking model. The proposed method can effectively prevent tracking module degradation introduced by target appearance changes. Extensive experiments conducted on public available data sets OTB-50/100 demonstrate that the proposed tracking method is robust to the background clutter and discriminative to the target scale variation. Both qualitative and quantitative results show the excellent performance against some state-of-the-art trackers.

Occlusion-Aware Correlation Particle Filter Target Tracking Based on RGBD Data

In recent decades, there have been considerable improvements in target-tracking algorithms. However, aspects such as target occlusion, scale variation, and illumination changes still present significant challenges to existing algorithms. In this paper, we describe an occlusion-aware correlation particle filter target-tracking method based on RGBD data. First, we derive a target occlusion judgment mechanism based on a depth image and the histogram of oriented gradients (HOG) feature. We then formulate the tracking mechanism for the target prediction–tracking–optimization–redetection process using a correlation maximum likelihood estimation particle filter algorithm. We propose an adaptive update strategy whereby the system saves a well-tracked model when no occlusion occurs, and then uses this saved model to replace poorly tracked models in the event of occlusion. Furthermore, we consider the scale variation and adjust the target size according to the depth image, but we leave the HOG feature vector dimension of the target area unchanged. Thus, the problems such as model offset, scale variation, and loss of features are corrected over time. The experimental results demonstrate that the proposed target-tracking algorithm can detect target occlusion and track targets well, requires fewer calculations to perform target prediction–tracking–optimization–redetection, reduces the impact of illumination changes, and achieves better real-time performance and accuracy than many existing algorithms.

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.

Robust Scale Adaptive Tracking by Combining Correlation Filters with Sequential Monte Carlo.

A robust and efficient object tracking algorithm is required in a variety of computer vision applications. Although various modern trackers have impressive performance, some challenges such as occlusion and target scale variation are still intractable, especially in the complex scenarios. This paper proposes a robust scale adaptive tracking algorithm to predict target scale by a sequential Monte Carlo method and determine the target location by the correlation filter simultaneously. By analyzing the response map of the target region, the completeness of the target can be measured by the peak-to-sidelobe rate (PSR), i.e., the lower the PSR, the more likely the target is being occluded. A strict template update strategy is designed to accommodate the appearance change and avoid template corruption. If the occlusion occurs, a retained scheme is allowed and the tracker refrains from drifting away. Additionally, the feature integration is incorporated to guarantee the robustness of the proposed approach. The experimental results show that our method outperforms other state-of-the-art trackers in terms of both the distance precision and overlap precision on the publicly available TB-50 dataset.

Real-Time Robust Tracking for Motion Blur and Fast Motion via Correlation Filters.

Visual tracking has extensive applications in intelligent monitoring and guidance systems. Among state-of-the-art tracking algorithms, Correlation Filter methods perform favorably in robustness, accuracy and speed. However, it also has shortcomings when dealing with pervasive target scale variation, motion blur and fast motion. In this paper we proposed a new real-time robust scheme based on Kernelized Correlation Filter (KCF) to significantly improve performance on motion blur and fast motion. By fusing KCF and STC trackers, our algorithm also solve the estimation of scale variation in many scenarios. We theoretically analyze the problem for CFs towards motions and utilize the point sharpness function of the target patch to evaluate the motion state of target. Then we set up an efficient scheme to handle the motion and scale variation without much time consuming. Our algorithm preserves the properties of KCF besides the ability to handle special scenarios. In the end extensive experimental results on benchmark of VOT datasets show our algorithm performs advantageously competed with the top-rank trackers.

Fast and robust object tracking via Accept–Reject color histogram-based method

We present a new framework for real-time tracking method of complex non-rigid objects. This new method successfully coped with camera motion, partial occlusions, and target scale variations. The shape of the object tracker is approximated by an ellipse and its appearance by histogram based features derived from local image properties. We use an efficient search scheme (Accept–Reject color histogram-based method (AR), using Bhattacharyya kernel as a similarity measure) to find the image region with a histogram most similar to the target of object tracker. In this paper, we address the problem of scale/shape adaptation and orientation changes of the target. The proposed approach is compared with recent state-of-the-art algorithms. Extensive experiments are performed to testify the proposed method and validate its robustness and effectiveness to track the scale and orientation changes of the target in real-time.

A coarse-to-fine model for airport detection from remote sensing images using target-oriented visual saliency and CRF

This paper presents a novel computational model to detect airports in optical remote sensing images (RSI). It works in a hierarchical architecture with a coarse layer and a fine layer. At the coarse layer, a target-oriented saliency model is built by combing the cues of contrast and line density to rapidly localize the airport candidate areas. Furthermore, at the fine layer, a learned condition random field (CRF) model is applied to each candidate area to perform the fine detection of the airport target. The CRF model is learned based on sparse features of local patches in a multi-scale structure and it also takes the contextual information of target into consideration. Therefore, its detection is more accurate and is robust to target scale variation. Comprehensive evaluations on RSI database from the Google Earth and comparisons with state-of-the-art approaches demonstrate the effectiveness of the proposed model.

Adaptive pyramid mean shift for global real-time visual tracking

Tracking objects in videos using the mean shift technique has attracted considerable attention. In this work, a novel approach for global target tracking based on mean shift technique is proposed. The proposed method represents the model and the candidate in terms of background weighted histogram and color weighted histogram, respectively, which can obtain precise object size adaptively with low computational complexity. To track targets whose displacements between two successive frames are relatively large, we implement the mean shift procedure via a coarse-to-fine way for global maximum seeking. This procedure is termed as adaptive pyramid mean shift, because it uses the pyramid analysis technique and can determine the pyramid level adaptively to decrease the number of iterations required to achieve convergence. Experimental results on various tracking videos and its application to a tracking and pointing subsystem show that the proposed method can successfully cope with different situations such as camera motion, camera vibration, camera zoom and focus, high-speed moving object tracking, partial occlusions, target scale variations, etc.

Kernel-based object tracking

A new approach toward target representation and localization, the central component in visual tracking of nonrigid objects, is proposed. The feature histogram-based target representations are regularized by spatial masking with an isotropic kernel. The masking induces spatially-smooth similarity functions suitable for gradient-based optimization, hence, the target localization problem can be formulated using the basin of attraction of the local maxima. We employ a metric derived from the Bhattacharyya coefficient as similarity measure, and use the mean shift procedure to perform the optimization. In the presented tracking examples, the new method successfully coped with camera motion, partial occlusions, clutter, and target scale variations. Integration with motion filters and data association techniques is also discussed. We describe only a few of the potential applications: exploitation of background information, Kalman tracking using motion models, and face tracking.