针对野外大视场、远距离、随机出现的特征相同多物体定位的实际需求，构建了基于双目立体视觉原理的三维坐标测量系统。为了高精度定位随机出现的特征相同多物体，需要正确匹配多物体同名点，因而提出了一种基于Delaunay三角剖分和仿射约束的立体匹配方法。利用具有抗仿射变换的仿射尺度不变特征变换(ASIFT)算法获得左右背景图像匹配点；通过Delaunay三角剖分算法对抽样后的匹配点生成三角网格，对左右图像每一对匹配三角形区域计算仿射矩阵；根据多物体同名点在不同匹配三角形内的分布，利用仿射约束实现对多物体同名点的匹配。实验结果表明，提出的方法对特征相同多物体同名点实现了快速高精度的匹配，多物体同名点提取以及实时匹配时间约为30 ms，满足相机25 frame/s的实时处理要求，较好地解决了超大三维空间中弧形坡面上随机出现的特征相同多物体同名点的匹配问题。

For practical demand of the localization of multiple random objects with large view filed, long distance and same features, a 3D coordinate measuring system is established based on the binocular stereo vision theory. To precisely position the multiple random objects with the same features, the multiple objects need matching correctly. An innovative method based on the Delaunay triangulation and affine constraint is proposed to achieve correct matching of the multiple objects with same features. The matching points on the background images are obtained with the affine scale-invariant feature transform (ASIFT) algorithm that has an anti-affine transformation. The Delaunay triangulation algorithm is used to generate triangular meshes by the seed points. The affine matrix of the triangular region is calculated by using vertexes of matched triangles. According to the distribution of object points in different matched triangles, the multiple objects with same features will be matched by the affine constraint. Experimental results show that the proposed method realizes the fast and efficient matching of multiple objects with same features. The time of object extraction and real-time matching is about 30 ms, which satisfies the requirement of 25 frame/s real-time processing for cameras. The proposed method solves the problem of matching of multiple objects with same features on the arc slope in large 3D space.