Object Detection Technology Based On Online Learning Method

Online learning for object detection is one of comparatively new topic in computer vision community. The main research content is designing intelligent object detection system which can self learning and improve its detection performance like human visual system. But it is difficult to find a universal model fitting to this research task. This thesis presents an online learning object detection method which combines tracking information. And the samples are acquired and labeled automatically when they are used to training the detector online. Our research focuses on improving the classification capability of online learning classify and labeling the online learned samples automatically.Adaptive cascade classifier algorithm is proposed to improve online learning algorithm. Cascade decision strategy is integrated with the online boosting procedure. The resulting system contains enough number of weak classifiers while keeping computation cost low. The cascade structure is learned and updated online. And the structure complexity can be increased adaptively when detection task gets more difficult. Experimental results show the method is more efficient and accurate.Next we propose a novel online learning framework for object detection in video sequences. At first, an off-line classifier is trained with a few labeled samples. And it was used to object detection in video sequences. Based on online learning algorithm, the detected objects will be used to train the classifier as new samples. Instead of using another detection algorithm to label the new sample automatically like other online learning framework, we ensure the correct label from tracking. This can greatly reduce the effort by labeler. A number of experimental results demonstrate the effectiveness of the method.Further, we present a particle filter tracking algorithm based on cues integration mechanism to improve sample labeling accuracy. The information along contours and object boundaries is used to the likelihood model of particle filter algorithm. This improved our online learning system.Moreover, we present a new boosting based learning method where the linear regress tree is used to automatically combine multiple features in weak leaning process. And the final strong classifier is composed of several weak linear regress trees. This leads to a better strong classifier, which consists of fewer weak classifiers and features. Recursive least square algorithm can be used to update the linear regress coefficients at each node in linear regress tree. So the online learning weak linear regress tree algorithm can be implemented.At last, we summarize the presented work. According to the imperfect aspects, we analyze and discuss the future work.