【作者】 王昆；

【导师】 付宇卓；

【作者基本信息】 上海交通大学 ， 软件工程， 2007， 硕士

【摘要】 视频序列中运动目标的检测是计算机视觉和图像编码研究领域的一个重要课题,在机器人导航、智能监视系统、交通监测、医学图像处理以及视频图像压缩和传输等领域都有广泛的应用。FPGA作为当今主流的大规模可编程专用集成电路,可以满足高速图像处理的需要。使用FPGA可以充分利用硬件上的并行性,从本质上改善图像处理的速度,使对大数据量的图像处理达到实时性。本文提出基于FPGA的运动目标检测系统,对以后算法的改进,输入输出图像大小的变化,图像采集和显示设备更换等都具有灵活性。本文对目前运动目标检测的主要算法研究分析,根据背景减法的适用环境和特点提出改进的W4运动检测算法。该算法具备背景减法的优点,并且克服了W4运动检测算法在环境变化较快或环境变化较频繁条件下对运动目标进行检测的局限性。本文首先在MATLAB中对改进的W4运动检测算法进行仿真,然后将算法移植到FPGA中实现。设计图像采集、图像检测和VGA显示等模块,完善运动目标检测系统。根据算法和运动目标检测系统的特点提出一种基于改进的W4算法的快速检测方法,该方法以块为单位进行运动目标检测,可以有效地提高图像处理的速度,使系统满足实时性要求。更多还原

【Abstract】 Motion objects detection is an important problem of computer vision and image coding. It is widely used in automaton navigation, intelligent video surveillance systems, traffic detection, medical image processing, video compression and transmission and so on. As a main large scale programmable specific integrated circuit, FPGA meets the requirements of high speed image processing, with the characteristics of simple algorithm realization, easy programming, good portability and inheritability. In this thesis, we develop a motion detection system based on FPGA. It is flexible for later changing of algorithm, image size and collection and display equipments.By analyzing some of the main algorithm of motion detection, especially background subtraction, a new motion detection algorithm called improved W4 is presented. The novel algorithm has the advantages of background subtraction and overcomes the shortages of W4 algorithm that can not efficiently detect moving object in frequent varying environmental weather condition.We use MATLAB to simulate the improved W4 algorithm, and then port it to the FPGA. We design image collection, motion detection and display modules to consummate the motion detection system. In order to accelerate it to be a real-time system, a fast algorithm which uses block instead of pixel in improved W4 algorithm is presented.更多还原