摘要
利用双目立体视觉技术对空间物体进行定位的一个难点在于特征点的选择和匹配,常用的方法是选择一幅图像中物体的形心为特征点,然后根据灰度或者颜色值在另外一幅中进行匹配。为了达到较好的精度,所用的匹配算法都比较复杂,并且当物体表面颜色分布均匀时错误匹配率较高。针对球形物体,本文提出基于形心几何匹配的定位方法,避开了常规匹配方法所面对的难题,经过试验验证,效果很好。
One problem of locating global object in the space using binocular stereo vision is which point should be picked as key point and how to match the key point in the two images. The common way is selecting the centroid as the key point and then matches it in the two images according to color values or gray values. The matching algorithms usually are complex and wrong results are got sometimes,especially when the color distribution in the surface of the object is symmetrical. In this paper,a new method is presented aiming at global object. Based on centroid geometric matching and it avoided the problem that the old methods had to meet with. It is proved effective by experiments.
出处
《现代电子技术》
2005年第17期78-80,共3页
Modern Electronics Technique
关键词
球形物体
双目立体视觉
定位
形心
global object
binocular stereo vision
locating
centroid
