立体视觉组网测量是三维测量得以实现的核心技术，在测量过程中，为了保证三维重构的全覆盖和重构精度，需要摄像机进行密集拍摄，因此产生了三维测量的速度慢，计算量较大等问题。针对上述问题，提出了一种多视点立体视觉测量网络组网方法。首先，通过SFM技术（运动恢复结构）获得待测物模型，建立椭球体基准坐标，估计相机与待测物的最佳距离，布置初始视点位置；然后，基于可视化约束条件对初始视点进行聚类分析和循环迭代，筛选出最少的相机数目实现全覆盖三维成像；最后，进行测量实验，布置以灯罩为待测物的椭球形测量网络，并在不同景深下进行相机数目、覆盖率和测量精度与球形测量网络的对比分析。实验结果表明：通过该方法最终迭代筛选出了22个视点，使得覆盖率达到100%，测量精度的标准差均值稳定到1.1 mm，测量效率较球形网络有明显提升，经三维重建出的灯罩效果图能够还原出待测物原貌，验证了所提出方法的可行性。

Stereo-vision network measurement was the core technology for 3D measurement. In order to ensure the full coverage and reconstruction accuracy of 3D reconstruction, the camera needed to be intensively shot during the measurement process, resulting in slow 3D measurement and large calculation.Therefore, a multi-view stereo-vision measurement network networking method was proposed to solve the above problems. Firstly, the object model was obtained by SFM technology (motion recovery structure), establishing the ellipsoid reference coordinates, estimating the optimal distance between the camera and the object to be tested, and arranging the initial viewpoint position. Secondly, the minimum number of cameras, to achieve full coverage 3D imaging, was filtered based on visual constraints to cluster analysis and loop iteration of the initial viewpoint. Finally, the measurement experiment was carried out, and the ellipsoidal measurement network with the lampshade as the object to be tested was arranged. The comparison of the number of cameras, coverage and measurement accuracy with the spherical measurement network was carried out at different depths of field. The experimental results show that 22 viewpoints are selected through the final iteration of the method, so that the coverage rate reached 100%. The standard deviation of the measurement accuracy was stable to 1.1 mm and the measurement efficiency was significantly improved compared with the spherical network. The original appearance of the object to be tested was restored through the 3D reconstruction of the lampshade rendering, which verified the feasibility of the proposed method.