1、视频图像中运动物体的检测 I 摘摘 要要 运动目标检测处于整个视觉检测系统的最底层，是各种后续高级处理如目标分 类、行为理解等的基础。运动目标检测是指从视频流中实时提取目标，一般是确定 目标所在区域和颜色特征等。目标检测的结果是一种“静态”目标前景目标， 由一些静态特征所描述。 在论文的第一部分介绍了运动目标检测的基本理论以及实现方法；在论文的第 二部分简述了模式识别，数学形态学以及边缘检测的概念，重点介绍了数学形态学 中的腐蚀和膨胀以及 canny 算子用于边缘检测的方法。 通过观察所需检测的图像的 性质，笔者首先使用连通性的概念，进行了算法一的尝试；在不成功后，笔者又利 用数学形态学和模式
2、识别结合的方法，在 canny 算子边缘检测的基础上，进行了算 法二的初步尝试，得出了检测结果。并通过对比试验，检测了算法的效果。 关键词关键词：运动目标检测，数学形态学，模式识别，边缘检测 视频图像中运动物体的检测 II Abstract The locomotion target detecting is placed in the understratum of the whole sense of vision detecting system, it is also the base of every kind of follow-up advanced dispose such a
3、s the target sorting, behavior comprehending and so on. The locomotion target detecting is pick-uping target in real time from the video stream; it is commonly making sure the region and color characteristic etc. Of the target. The result of the target detecting is a kind of “ static “ target foregr
4、ound target, being described by some static characteristic. The basic theoretics and realization method of the locomotion target detecting is introduced in the first part of the thesis; in the second part of the thesis it is simply depicting the conception of pattern identify, mathematical morphology and the edge deetecting,it is mostly introduced the method of the eroding and dilating of mathematical morphology, and the edge detecting using canny operator. Acrossing the character of the
