摘 要随着科技的进步，智能视频监控系统的应用越来越广泛，智能监控系统一般的处理流程首先进行运动目标检测、随后进行运动目标跟踪、然后进行特征提取和最后进行行为分类和识别。运动目标检测是智能监控系统整个过程中首要的步骤，其结果的好坏对其他的流程有着重要的影响。87152

本文详细的介绍了目前常用的一些运动目标检测算法。具体来说，介绍了光流法、帧间差分法和三帧差分法，然后通过实验数据对这些算法进行了比较，从而深刻学习了这些算法，了解了这些算法各个方面的优劣。

本文首先介绍了在对视频序列处理之前在预处理阶段对图像进行一般需要先对图像进行灰度化处理，从而减少图像处理工作量，提高处理效率，并对图像进行均值滤波去噪，中值滤波去噪，介绍并实现了这两种去噪方法，并对这两种滤波去噪方法对加入了椒盐噪声的图像进行了处理，从而得出结论中值滤波算法比均值滤波去噪算法能更好的去除差异较大的噪声。

随后本文对常见的运动目标检测算法，如光流法，帧间差分法和三帧差分法进行了基本原理介绍与算法流程分析，主要介绍了帧间差分法及其改进算法三帧差分法，并实现了这两种算法，并在运动目标移动较快与较慢时进行了图像处理结果的比较，从而证明三帧差分法改进的有效性。

最后本文对经过算法识别的运动目标进行了数学形态学处理，通过膨胀运算填补空洞，通过腐蚀运算获取更为精确的运动目标边缘，通过开运算使边缘更为圆滑，通过闭运算消除内部空洞，选用适当的数学形态学处理可以使得检测结果更为理想 。

毕业论文关键词：概率模型；运动目标检测；图像去噪；

Abstract With the progress of science and technology, the application of intelligent video surveillance system is more and more widely, the intelligent monitoring system of general process for moving target detection, first then to moving target tracking, and then to feature extraction and finally behavior classification and recognition。 Moving target detection is the first step in the process of intelligent monitoring system, the result is good or bad for other process has important influence。

In this paper in detail introduces the commonly used some of the moving target detection algorithm。 Specifically, this paper introduces the optical flow, interframe difference and three frame difference method, and then through the experimental data of these algorithms are compared, so as to deeply study the algorithm, to understand the advantages and disadvantages of these algorithms all aspects。

Paper in the second chapter on the video sequence processing before the pretreatment stage for the general need to gray-scale image processing, thus reduce the workload of image processing, improve the efficiency of processing, and median filter for image denoising and median filtering de-noising, introduced and implemented both denoising method, and the two filtering denoising method to add the salt and pepper noise image processing, and thus concluded that median filtering algorithm is better than average filtering denoising algorithm can remove the noise of the differences。

Later in this paper, the common motion target detection algorithm, such as optical flow, interframe difference method and three frame difference method to introduce the basic principle and algorithm of process analysis, this paper introduces the frame difference method between three frame difference method and its improved algorithm, and realizes the two algorithms, and the moving target with faster and slower when comparing the results of the image processing, thus to demonstrate the effectiveness of three frame difference method is improved。