In order to resolve the mismatching problems caused by the large change in the depth of the obj ects or the complex surface shape in the field of PMP visual measurement,this paper proposed a stereo matching method based on region segmentation and parallax correction.The basic principle of four-step phase shift method and the multi-frequency heterodyne were investigated,and stereo rec-tification method in OpenCV software package was used to rectify the raster image.The field of view of the cameras was analyzed according to the characteristics of the PMP vision measurement system. Using the characteristics of the phase information in the field of view and the parallax constraint,the principle and the method of region segmentation were proposed based on the mixed threshold.Fur-thermore,the valid parallax maps were obtained by parallax solving and parallax correction,and sev-eral test experiments were carried out to demonstrate the effectiveness of the proposed stereo matc-hing method in PMP visual measurement system.%针对P MP 视觉测量中物体深度变化大或表面形状复杂时出现的误匹配问题，提出一种基于区域分割和视差修正的立体匹配方法。研究了四步相移法和多频外差相移的基本原理，并采用基于OpenCV的标定立体校正方法对光栅图像进行了校正；结合 PMP 视觉测量系统的特点对相机视场进行分析，根据视场各区域的相位信息特点以及视差约束，提出基于混合阈值的区域分割原则和实现方法；应用视差求解和视差校正获得有效的视差图，并在自主研发的 P MP 视觉测量系统中验证了立体匹配方法的有效性。
展开▼
