| The technology of video based object detection and tracking is one of the hotspots in the field of computer vision. Human beings sense the environment through sensory organs, and vision is our main information source. The aim of computer vision is to let computers have the ability of learning on surrounding information. There is more useful information in motion sequences than static images. Extraction of moving objects is to find differential objects in motion sequences and mark them. Object tracking is a new research topic with the developing of digital video technology. It has been widely used both in military and civil fields.The fundamental theories and key technologies of video-based detection and tracking for moving objects are mainly discussed in this thesis. The following topics are researched in details, such as detection and extraction of moving objects under a background model, detection of shadow of moving objects and tracking of moving objects and so on.On the research of object detection, current theories of object detection and extraction algorithms are summarized. Temporal Differential method and Adaptive Background Subtraction method are compared with each other through experimental results. Based on that, we proposed a new detection algorithm combining background subtraction and time difference. The experimental results show that the improved scheme can satisfy the need of system and improve the accuracy of detected result.On the research of object tracking, firstly we researched several main object features and matching algorithms. Secondly we researched the Mean-Shift Algorithm (MSA) based on object feature of color histogram and proposed an improved algorithm based on MSA. Comparing the results of experiment of MSA and the improved algorithm, the experimenting results of the improved algorithm is better than MSA because of using the spatial information of objects. The improved algorithm increased reliability and validity of tracking results... |
PDF Full Text Request