Dawei Liang

A scheme for ball detection and tracking in broadcast soccer video

In this paper we propose a scheme for ball detection and tracking in broadcast soccer video. There are two alternate procedures in the scheme: ball detection and ball tracking. In ball detection procedure, ball candidates are first extracted from several consecutive frames using color, shape, and size cues. Then a weighted graph is constructed, with each node representing a candidate and each edge linking two candidates in adjacent frames. Finally, Viterbi algorithm is employed to extract the optimal path as ball’s locations. In ball tracking procedure, Kalman filter based template matching is utilized to track the ball in subsequent frames. Kalman filter and the template are initialized using detection results. In each tracking step, ball location is verified to update the template and to guide possible ball re-detection. Experimental results demonstrate that the proposed scheme is promising.

Improving particle filter with support vector regression for efficient visual tracking

Particle filter is a powerful visual tracking tool based on sequential Monte Carlo framework, and it needs large numbers of samples to properly approximate the posterior density of the state evolution. However, its efficiency degenerates if too many samples are applied. In this paper, an improved particle filter is proposed by integrating support vector regression into sequential Monte Carlo framework to enhance the performance of particle filter with small sample set. The proposed particle filter utilizes an SVR based re-weighting scheme to re-approximate the posterior density and avoid sample impoverishment. Firstly, a regression function is obtained by support vector regression method over the weighted sample set. Then, each sample is re-weighted via the regression function. Finally, ameliorative posterior density of the state is re-approximated to maintain the effectiveness and diversity of samples. Experimental results demonstrate that the proposed particle filter improves the efficiency of tracking system effectively and outperforms classical particle filter.

Object tracking using incremental 2D-LDA learning and Bayes inference

The appearances of the tracked object and its surrounding background usually change during tracking. As for tracking methods using subspace analysis, fixed subspace basis tends to cause tracking failure. In this paper, a novel tracking method is proposed by using incremental 2D-LDA learning and Bayes inference. Incremental 2D-LDA formulates object tracking as online classification between foreground and background. It updates the row- or/and column- projected matrix efficiently. Based on the current object location and the prior knowledge, the possible locations of the object (candidates) in the next frame are predicted using simple sampling method. Applying 2D-LDA projection matrix and Bayes inference, candidate that maximizes the posterior probability is selected as the target object. Moreover, informative background samples are selected to update the subspace basis. Experiments are performed on image sequences with the objects appearance variations due to pose, lighting, etc. We also make comparison to incremental 2D-PCA and incremental FDA. The experimental results demonstrate that the proposed method is efficient and outperforms both the compared methods.

Mean-Shift Blob Tracking with Adaptive Feature Selection and Scale Adaptation

When the appearances of the tracked object and surrounding background change during tracking, fixed feature space tends to cause tracking failure. To address this problem, we propose a method to embed adaptive feature selection into mean shift tracking framework. From a feature set, the most discriminative features are selected after ranking these features based on their Bayes error rates, which are estimated from object and background samples. For the selected features, a criterion is proposed to evaluate their stability for tracking and to guide feature reselection. The selected features are used to generate a weight image, in which mean shift is employed to locate the object. Moreover, a simple yet effective scale adaptation method is proposed to deal with object changing in size. Experiments on several video sequences show the effectiveness of the proposed method.

Video2Cartoon: A System for Converting Broadcast Soccer Video into 3D Cartoon Animation

A system for converting broadcast soccer video into 3D cartoon animation is proposed. The system can provide viewers with a new experience that can not be acquired from the original soccer video, by taking advantage of computer vision and computer graphics techniques. Firstly, computer vision techniques are employed to estimate the 3D positions of players and the ball. Then, players, the ball and the playfield are modeled by computer graphics techniques. Finally, 3D cartoon animation is generated based on the extracted 3D information and the pre-constructed player motion database. The system allows users to watch the game at arbitrary viewpoint using a virtual camera. On the other hand, content service providers can distribute the generated cartoon animation on their Web portals or to mobile devices, for consumers to better enjoy the game.

Video2Cartoon: generating 3D cartoon from broadcast soccer video

In this demonstration, a prototype system for generating 3D cartoon from broadcast soccer video is proposed. This system takes advantage of computer vision (CV) and computer graphics (CG) techniques to provide users new experience that can not be obtained from original video. Firstly, it uses CV techniques to obtain 3D positions of the players and ball. Then, CG techniques are applied to model the playfield, players, and ball. Finally, 3D cartoon is generated. Our system allows users to watch the game at any point of view using a 3D viewer based on OpenGL.

DRM: dynamic region matching for image retrieval using probabilistic fuzzy matching and boosting feature selection

This paper considers the semantic gap in content-based image retrieval from two aspects: (1) irrelevant visual contents (e.g. background) scatter the mapping from image to human perception; (2) unsupervised feature extraction and similarity ranking method can not accurately reveal users’ image perception. This paper proposes a novel region-based retrieval framework—dynamic region matching (DRM) to bridge the semantic gap. (1) To address the first issue, a probabilistic fuzzy region matching algorithm is adopted to retrieve and match images precisely at object level, which copes with the problem of inaccurate segmentation. (2) To address the second issue, a “FeatureBoost” algorithm is proposed to construct an effective “eigen” feature set in relevance feedback (RF) process. And the significance of each region is dynamically updated in RF learning to automatically capture users’ region of interest (ROI). (3) User’s retrieval purpose is predicted using a novel log-learning algorithm, which predicts users’ retrieval target in the feature space using the accumulated user operations. Extensive experiments have been conducted on Corel image database with over 10,000 images. The promising experimental results reveal the effectiveness of our scheme in bridging the semantic gap.

Online selection of discriminative features using bayes error rate for visual tracking

Online feature selection using Bayes error rate is proposed to address visual tracking problem, where the appearances of the tracked object and scene background change during tracking. Given likelihood functions of the object and background with respect to a feature, Bayes error rate is a natural way to evaluate discriminating power of the feature. From previous frame, object and background pixels are sampled to estimate likelihood functions of each color feature in the form of histogram. Then, all features are ranked according to their Bayes error rate. And the top N features with the smallest Bayes error rate are selected to generate a weight map for current frame, where mean shift is employed to find the local mode, and hence, the location of the object. Experimental results on real image sequences demonstrate the effectiveness of the proposed approach.

Novel observation model for probabilistic object tracking

Treating visual object tracking as foreground and background classification problem has attracted much attention in the past decade. Most methods adopt mean shift or brute force search to perform object tracking on the generated probability map, which is obtained from the classification results; however, performing probabilistic object tracking on the probability map is almost unexplored. This paper proposes a novel observation model which is suitable to perform this task. The observation model considers both region and boundary cues on the probability map, and can be computed very efficiently by using the integral image data structure. Extensive experiments are carried out on several challenging image sequences, which include abrupt motion change, background clutter, partial occlusion, and significant appearance change. Quantitative experiments are further performed with several related trackers on a public benchmark dataset. The experimental results demonstrate the effectiveness of the proposed approach.

Self-calibration based 3d information extraction and application in broadcast soccer video

This paper proposes a new method based on self-calibration to estimate the ball’s 3D position in broadcast soccer video. According to the physical limitation, the ball’s 3D position is estimated through the camera position and the ball’s virtual shadow, which is the point of intersection between the playfield and the line through the camera’s optical center and the ball. First, the virtual shadow is computed by the homography between playfield and image plane. For the image having enough corresponding points, the map is determined directly; for those images not having enough these points, their homographies are estimated through global motion estimation. Then, based on self-calibrating for rotating and zooming camera, and the homography, the camera’s position in the playfield is estimated. Experiments show that the proposed method can extract ball’s 3D position information without referring to other object with assuming height and obtain promising results.