视频目标侦测与智能监控技术.docVIP

视频目标侦测与智能监控技术及其IP设计 信息科学与工程学院2008级集成电路设计与集成系统专业 李江涛 指导教师 陈江华HDL的形式硬件实现其IP软核模块。将和基于Haar特征的AdaBoost目标识别算法结合起来。Android嵌入式平台了算法。分辨率800x480的视频进行实时处理，帧率在20fps以上。在软件实现快速可靠的算法基础上，HDL的形式设计了识别率高、资源消耗较少的硬件算法IP核，完成了算法的设计与验证，并对其进行了性能分析。经验证，分辨率800x480的视频流，25MHz主时钟下满足嵌入式的实时视频的应用。AdaBoost算法；IP核Abstract： Based on object recognition in video streaming and video motion detection technology, the intelligent monitoring technology leads the direction of video surveillance system in the future. At the same time, video object detection technology is widely used in many other areas. This project is to design a stable and reliable algorithm about video motion detection and object recognition, and implement it in the form of synthesizable HDL as an IP core module. It is an innovation to combine video motion detection and the AdaBoost object detection Algorithm based on Haar features together. This will complement its weakness each other. I use ARM-Android embedded platform to develop and debug the video motion detection and the AdaBoost object detection algorithm. This algorithm runs up to 20fps at a 1.2GHz CortexA9 ARM processer. Based on the achievement of the fast and reliable algorithm, I have designed a low resource consumption algorithm implemented by hardware. We have completed the design and validation of the IP core module, as well as its performance analysis. For a video stream with a resolution of 800x480 and a frame rate of 30 fps, a 25MHz clock is enough to meet the need of real-time video object detection. Keywords: video stream; target detection; AdaBoost algorithm; IP core 一、绪 论 1.1 课题背景及意义 1.1.1 课题背景 视频移动侦测技术与视频目标识别技术是视频信号处理、分析应用的一个重要领域。基于移动侦测技术和视频目标识别技术的智能监控系统（IVS)是未来监控的发展方向。 视频移动侦测技术（VMD），就是使用计算机图像处理技术，将场景中运动的目标提取追踪。视频目标识别技术则是通过检测不同的特征库识别指定的目标。 1.1.2本课题的研究意义 智能视频系统可以将视频的记录与分析交由计算机，使监控工作更加高效。视频分析技术将大幅提升安防工作的效率与质量，使得整个社会更加安定,和谐。 使用嵌入式实现智能监控系统比基于PC的设备在可扩展性，功耗等方面优势明显。此外，视频处理需要很高的处理器性能。使用高性能的ARM处理器和DSP将带来成本、功耗的提升，因而使用硬件实现相关算法具有重要意义。 1.2 相关问题的研究现状 1.2.1 研究现状 关于视频移动侦测的问题人们