Research On Classification Of Moving Objects In Video Sequence

With the rapid development of computer hardware and software, Video-based motion analysis increasingly attracts people’s attention. Video-based motion analysis aims at detecting, tracking and identifying moving objects, and more generally, understanding objects behaviors through analysis and processing image sequences with moving objects. Objects classification is an important aspect of video-based motion analysis whose research content is to classify moving objects into semantically meaningful categories. Classification is significant for the development of automatic video comprehension and is the foundation of behavior analysis and understanding.The moving object classification of normal outdoor scenes based on static odd-camera is studied in this paper. On the basis of summarizing and analyzing actuality research and algorithms both here and aboard, an object classification algorithm based on shape features and time context information is proposed. With support vector machines as the classifier, it classes the objects detected in video into four familiar sorts:person, vehicle(including car, van, truck), crowd, bicycle(including bicycle, motorcycle, electric cars, we unified classify in bicycle for convenience of description). The main work is as follows:Firstly, video images are pretreated through segmenting regions corresponding to objects then tracking them after mask fill background subtraction and with different regions of the right value to update background. Combine the FNCC with HSV algorithms to eliminate the shadow of the object so as to subtract the basic outline of the object. In the target tracking, using MeanShift iteration and multi-target algorithm based on target detection algorithm to track the target.Secondly, extracting the features of moving objects based on the previous work. After studying the current image characterization method. This paper choose the shape features and velocity of the objects as the objects’features, which can fit for the object is blocked or includes some shadow.Then, multi-class SVM based on small sample is selected as the classifier. Labeled samples are used to train the classifier. The trained classifier can be used to class unknown object samples. The problem of multi-class non-linear object classification can be solved better.At last. Some means which improve the performance of object classification are discussed. The time complexity of classification algorithm is lower and the real-time of the system is improved with the effective area detection and interval frame classification. The accuracy of classification is improved after using temporal consistency.Simulation results show that this method can class person, vehicle, crowd, bicycle effectively.