The Dm642 Video Processing And Transmission Systems Software Design And Pedestrian Target Detection, Based On Video Content

Video surveillance of digital and networking technology is the inevitable trend of development. TI DSP-DM642 has powerful video processing capabilities and a lot of external interfaces. Taking it as the main chip, the embedded video surveillance system could be small in size and functional fully. In this paper, the video processing and transmission system has the functions such as multi-channel data acquisition, data compression, local storage, networking transmission and so on. These operations are all controllable. The first part of this paper mainly introduces the system's software design including software architecture, functional block design and some hardware driver programming.Human detection is a great challenge in the fields of target detection of machine vision. It also has great practical value. The latter part of this paper focuses on pedestrian detection in video content. The anther goal of this paper is to detect pedestrian in complex and cluttered scenes. And there is no any assumption on features of texture, color or edges of pedestrian or background. We adopt Haar wavelet for image feature extraction and the Support Vector Machine for classification. The entire detection system is composed of two parts. The first part is called full-body detection system that can check out the windows which may have pedestrian. The second part is a component-based detection system which is applied to the former windows for head-and-shoulder, left arm, right arm and lower body detection, then it makes the final decision whether the windows containing pedestrian.