发布时间 : 星期日 文章毕业论文 - 视频图像中运动物体的检测更新完毕开始阅读573dc5aa33d4b14e84246828
视频图像中运动物体的检测
摘 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：运动目标检测，数学形态学，模式识别，边缘检测
I
视频图像中运动物体的检测
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 as 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 \—— foreground 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 observing a kind of picture needing detecting, the author firstly uses the conception of connectedness, and proceed according to the method of connectedness target detecting, but it is unsuccessful, the author according the method of uniting mathematical morphology and pattern identify, on the basis of edge detecting using canny operator, proceed according to the method of difference of two frames threshold value and educe the detecting result. At last , the author according contrast experimentation, detect effect of the arithmetic.
Key Words: target detecting, mathematical morphology, pattern identify,
edge detecting
II
视频图像中运动物体的检测
引 言
在当今信息时代，社会的发展突飞猛进，各行各业都离不开信息，尤其是图像信息。图像信息处理作为一门跨学科的前沿科技已经广泛的应用在各个领域。运动目标检测是图像信息处理中的重要课题，运动目标检测的结果,常常是下一步的目标跟踪、模式识别、图像理解等高级后处理的输入图像。在许多场合,比如说交通流量的监测等,我们往往对运动的物体更感兴趣。因此研究只对运动目标敏感的检测与跟踪系统是很有意义的。
纵观全文，本论文主要完成了以下工作：首先，介绍了边缘检测的基本概念，边缘检测算子的处理，并特别着重介绍了canny算子；其次，介绍了数学形态学和模式识别，特别详细介绍了腐蚀，膨胀，连通性和阈值分割的概念和算法；最后，介绍了作者所使用的两种算法，并做出了对比实验。
III
视频图像中运动物体的检测
目录
摘 要 ................................................................ I Abstract ............................................................. II 引 言 ............................................................... III 第一章
绪言 ....................................................... １ 1.1 引 言 ..................................................... １ 1.2运动目标检测的主要方法 ..................................... ２ 1.3 论文的主要内容 ............................................ ３
第二章 视频图像中运动目标检测 ........................................ ４
2.1数学形态学 ................................................. ４ 2.1.1腐蚀 ..................................................... ４ 2.1.2膨胀 ..................................................... ６ 2.1.3连通性 ................................................... ７ 2.2边缘检测 ................................................... ８ 2.2.1Robert算子 ............................................... ９ 2.2.2Sobel算子 ................................................ ９ 2.2.3Prewitt算子 .............................................. ９ 2.2.4LOG算子 ................................................ １０ 2.2.5canny算子 .............................................. １０ 2.3基于阈值的图像分割 ....................................... １３ 2.4视频图像运动目标检测算法 ................................. １５ 2.4.1基于连通性运动目标检测 ................................. １５ 2.4.1.1算法框图 ............................................. １５ 2.4.1.2算法描述 ............................................. １７ 2.4.2基于帧间差阈值法的运动目标检测 ......................... １８ 2.4.2.1算法框图 ............................................. １８ 2.4.2.2算法描述 ............................................. ２０
第三章 仿真实验 ................................................... ２２
3.1单运动目标检测 ........................................... ２２ 3.2多运动目标检测 ........................................... ２６
第四章 结束语 .................................................... ２９ 参考文献 ........................................................... ３０ 附录................................................................ ３１
源程序 ...................................................... ３１
致谢 ................................................................................................................................... ３７
IV