昆虫运动图像分割中的阴影消除算法  

A Shadow Elimination Algorithm in Moving Insect Image Segmentation

作　　者：许立民[1] 何斌[1] 黎明和[1] 

机构地区：[1]同济大学电子与信息工程学院,上海201804

出　　处：《计算机与数字工程》2009年第3期161-163,171,共4页Computer & Digital Engineering

基　　金：国家自然科学基金新型湿吸机器人驱动原理研究(编号:50405045)资助;国家高技术研究发展计划(863计划)仿生湿吸微型机器人项目(编号:2007AA04Z253)资助

摘　　要：昆虫运动图像分割是昆虫步态提取、跟踪及分析的基础,根据昆虫运动图像中阴影点灰度小和色度保持稳定的特点提出一种适合室内运动检测的图像分割及阴影消除方法。先将当前帧图像与当前的背景图像对应像素点的像素值相除,比值小于1的点可能包括阴影和运动目标;然后计算比值小于1的点在RGB空间中的向量与对应背景点向量的夹角余弦。因为阴影色度保持不变,其夹角为0或极小,所以可将余弦接近1的点判为阴影。实验表明该方法简单有效,提取出来的昆虫前景图轮廓清晰,特征部分保持完整,为昆虫步态检测、分析及机器人步态规划提供支持。Moving image segmentation is foundation of extraction and analysis of gait of insect. In this paper, a shadow - eliminating algorithm for indoor moving detection based on the stability of shadow and the gray of shadow being less is presented. Firstly, the current frame image is divided by the current background image. The pixel whose ratio is less than 1 may be belong to shadow or motion object. Next, the cosine of the angle between vectors of the pixels whose ratio is less than 1 and corresponding pixels in background image in the RGB color space. The angles of shadow are or close to 0＊ because of the chrominance of shadow being stability, so the pixel can be detected a shadow dot when its cosine close to or is 1. The results of experiment show that the extracted foreground images of insect are distinct, and the algorithm is simple and effective for extraction and analysis of gait of insect.

关 键 词：图像分割 RGB颜色空间 运动检测 阴影消除 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]