Abstract: In order to extract the edge characteristics of the sole, a binocular stereo vision measurement method is proposed, in which light source is placed in the middle of sole, so as to increase internal and external contrast ratio of the edge
in captured images. As for feature matching at the?heel and toe cap, a stereo matching method is proposed based on extreme constraint. This algorithm is composed of the following three steps: a canny operator is applied to the image to extract edges. The algorithm fits a three time polynomial curve on edge points in the heel and toe cap, and obtains the edge points nearest from curve extreme point as feature points. The other edge points are matched using the average length. An experiment on sole with binocular 3D measurement method shows that adding source in the middle of sole can make the object edge clear, convenient for edge extraction. The 3D edge line of the sole feature obtained by the above matching algorithm is complete and correct, mismatching is restrained effectively.

Key words: binocular stereo vision, edge stereo matching, three type fitting；extreme constraint, light source

摘要： 为了提取鞋底边缘特征，提出了一种中间加光源的双目立体视觉测量方法，增大边缘内外的对比度。针对鞋头和鞋跟特征匹配，提出了基于极值约束的边缘立体匹配方法。该方法分三步：采用Canny算子提取左右图像的边缘；对鞋跟和鞋头部分的边缘点利用最小二乘三项式曲线拟合，求取边缘点上与曲线极值点最近的点作为极值特征点；利用长度均分法进行其他边缘点的匹配。对鞋底进行双目三维测量实验，结果表明，中间加光源可使得图像边缘清晰，便于边缘提取；提出的匹配方法获得的鞋底边缘三维数据完整、正确，有效地抑制了匹配错误。

关键词: 双目立体视觉, 边缘立体匹配, 三项式拟合, 极值约束, 光源