摘要
提出了一种基于立体视觉的物体边缘检测的方法。先对立体图像对进行基于图割的立体匹配方法求取场景的视差图,然后再用Canny的边缘检测方法对视差图进行边缘检测。立体视觉方法有效解决了单目视觉检测方法中的一些难点,利用了物体在空间的深度信息,对复杂背景下的物体和具有复杂纹理物体的边缘检测有很高的鲁棒性。实验结果表明该边缘检测方法优于传统的单目视觉边缘检测方法。
A novel edge detection algorithm based on stereo vision is proposed. At first, the paralicx map is obtained by stereo matching method via graph cuts, then the edge detection map is obtained from the parallex map by adopting Canny edge detection algorithm. The stereo algorithm can solve the problem effectively which cannot be solved by the monocular method. With the depth information of objects, the edge of objects in the complex background and objects with complex texture can be detected with high robustness. The experimental results prove its superiority over the traditional monocular edge detection method.
出处
《光学与光电技术》
2008年第4期71-74,共4页
Optics & Optoelectronic Technology
关键词
边缘检测
立体视觉
视差
立体匹配
edge detection
stereo vision
parallex
stereo matching
