基于视差与基线距相关三目立体匹配法

采用统计过程控制方法对二维图像特征点区域定位，并提取二维图像特征点。避免了提取二维图像特征点时，根据被处理图像的先验信息，利用试探方法确定阈值的局限性。在立体匹配时，将灰度相关系数小于最大灰度相关系数一定范围内的特征点作为灰度相关复峰初始匹配特征点集合。根据由正确匹配特征点组成的视差矩阵与对应的基线距矩阵存在极大相关性，从灰度相关复峰初始匹配特征点集合中确定唯一匹配特征点。通过对外形复杂的实际物体及已知精确三维坐标的标准工件的三维重建，证实了文中所提方法的有效性和可靠性。

Trinocular stereo matching algorithm based on correlation between disparities and baseline

After the area of feature points in 2-D image is located by statistical process control method,the feature points can be extracted easily.The limitation in determining threshold value with tentative algorithm according to some prior information on the image being processed can be avoided.During stereo matching, the feature points whose correlation coefficients are within certain range less than the maximal coefficient can be taken as an original matching feature point set.There is maximal correlation between the disparity matrix consisting of correct matching feature points and its corresponding baseline matrix.Based on this,by computing the correlations between disparity and the baseline matrices,the unique matching feature point can be obtained from the original matching feature point set.The images of an object with complex shape and a standard work-piece with given accurate 3-D coordinates were restored in 3-D form,which proves that the proposed algorithm is valid and reliable.