Bin Zheng

Bio: Bin Zheng is an academic researcher from Beijing Institute of Technology. The author has contributed to research in topics: Viola–Jones object detection framework & Video tracking. The author has an hindex of 2, co-authored 2 publications receiving 13 citations.

Object Tracking Algorithm Based on Combination of Dynamic Template Matching and Kalman Filter

Abstract: The moving object detection is a prerequisite and difficult point to realize tracking in the video tracking system. In order to detect moving object effectively, an object tracking algorithm is proposed based on combination of dynamic template matching and Kalman filter. First, get the area of the moving object by using inter-frame difference method and extract the SIFT feature points. Then, find the location of the candidate object that is most matched with the object template in the search area by Kalman filter and match it with the object template in the current frame. Finally, the feature points' loss rate will serve as the limited threshold, and we update template according to dynamic template updating strategy. By the number of the frames of the target's matching failures we determine whether the moving target is disappeared. Several experiments of the object tracking show that the approach is accurate and fast, and it has a better robust performance during the attitude changing, the size changing and the shelter instance.

An Adaptive Object Detection Scope Algorithm Based on SIFT

Abstract: For camera movement causes moving objects detecting and tracking problems under complex background, we propose an adaptive object detection scope algorithm based on SIFT features. Firstly, let camera stationary and obtain three images to detect the moving object by using three-frame-difference method, then extract the object SIFT features. Secondly, according to the location and displacement of the object in the dynamic background, we determine the detection scope which matches the object well and obtain the minimum rectangle which can surround the right matching points in the detection scope, and then update the object template. The algorithm avoids the analysis of the complex relative motion between the object and the background, and reduces mismatch points and the calculation amount. This algorithm can quickly and accurately track the object without occlusion, and performs robust in small occlusion case.

Target detection in SAR images using SIFT

Abstract: Target detection in synthetic aperture radar (SAR) images which are affected by speckle noise is a challenging task. An algorithm for automatic target detection in SAR images is proposed in this research work. In the first step, moving and stationary target acquisition and recognition (MSTAR) images are segmented and passed through multiple preprocessing stages (histogram equalization, dilation, position normalization). In the next step, feature extraction based on SIFT is performed. The extracted features from testing images are matched with the features extracted from training images. Thus, the classification of the targets is performed. The results obtained and the comparison with existing algorithms, both are sufficient enough to prove that the proposed algorithm is robust and effective.

Vision Based Multiple Target Tracking Using Recursive RANSAC

Abstract: Vision Based Multiple Target Tracking Using Recursive RANSAC James Kyle Ingersoll Department of Mechanical Engineering, BYU Master of Science In this thesis, the Recursive-Random Sample Consensus (R-RANSAC) multiple target tracking (MTT) algorithm is further developed and applied to video taken from static platforms. Development of R-RANSAC is primarily focused in three areas: data association, the ability to track maneuvering objects, and track management. The probabilistic data association (PDA) filter performs very well in the R-RANSAC framework and adds minimal computation cost over less sophisticated methods. The interacting multiple models (IMM) filter as well as higher-order linear models are incorporated into R-RANSAC to improve tracking of highly maneuverable targets. An effective track labeling system, a more intuitive track merging criteria, and other improvements were made to the track management system of R-RANSAC. R-RANSAC is shown to be a modular algorithm capable of incorporating the best features of competing MTT algorithms. A comprehensive comparison with the Gaussian mixture probability hypothesis density (GM-PHD) filter was conducted using pseudo-aerial videos of vehicles and pedestrians. R-RANSAC maintains superior track continuity, especially in cases of interacting and occluded targets, and has fewer missed detections when compared with the GM-PHD filter. The two algorithms perform similarly in terms of the number of false positives and tracking precision. The concept of a feedback loop between the tracker and sensor processing modules is extensively explored; the output tracks from R-RANSAC are used to inform how video processing is performed. We are able to indefinitely detect stationary objects by zeroing out the background update rate of target-associated pixels in a Gaussian mixture models (GMM) foreground detector. False positive foreground detections are eliminated with a minimum blob area threshold, a ghost suppression algorithm, and judicious tuning of the R-RANSAC parameters. The ability to detect stationary targets also allows R-RANSAC to be applied to a class of problems known as stationary object detection. Additionally, moving camera foreground detection techniques are applied to the static camera case in order to produce measurements with a velocity component; this is accomplished by using sequential-RANSAC to cluster optical flow vectors of FAST feature pairs. This further improves R-RANSAC’s track continuity, especially with interacting targets. Finally, a hybrid algorithm composed of R-RANSAC and the Sequence Model (SM), a machine learner, is presented. The SM learns sequences of target locations and is able to assist in data association once properly trained. In simulation, we demonstrate the SM’s ability to significantly improve tracking performance in situations with infrequent measurement updates and a high proportion of clutter measurements.

Object detection on robot operation system

Abstract: Object detection is a hot spot of the research in computer vision since many applications require the determination of the object location. There are many object detection methods based on feature matching methods. In this paper, we locate object on robot operation system. The SIFT keypoints of the template and test images are extracted at first. Then, the matching method is proposed to find the template image which is closest to the test image. The matching method is applied on the closest template image and the test images. Finally, we use affine transformation to get the rectangle represents the location of the object. Our system is a robot arm based on robot operation system(ROS). This robot arm consists of six degrees of freedom. In the experiment, we represent a result of object result using the images obtained form the camera installed on the robot arm. Our system is helpful to locate object on robot operation system.

Visual object tracking using improved Mean Shift algorithm

Abstract: Visual object tracking is one of many important applications for surveillance systems. The issues for visual object tracking are robustness from background interference, scaling and occlusion detection. In this paper, visual object tracking using improved Mean Shift algorithm is proposed. Mean Shift algorithm is used to obtain center object target for tracking. Corrected Background Weighted Histogram is added in target model to reduce background interference. Then, Scale adaptive methods is added in Mean Shift for scaling. Occlusion detection is handled by scaled Normalized Cross Correlation. The results prove that the proposed method is robust from noise background, scaling and occlusion detection.

Comparative survey: People detection, tracking and multi-sensor Fusion in a video sequence

Abstract: Tracking people in a video sequence is one of the fields of interest in computer vision. It has broad applications in motion capture and surveillance. However, due to the complexity of human dynamic structure, detecting and tracking are not straightforward. Consequently, different detection and tracking techniques with different applications and performance have been developed. To minimize the noise between the prediction and measurement during tracking, Kalman filter has been used as a filtering technique. At the same time, in most cases, detection and tracking results from a single sensor is not enough to detect and track a person. To avoid this problem, using a multi-sensor fusion technique is indispensable. In this paper, a comparative survey of detection, tracking and multi-sensor fusion methods are presented.