In response to the demand for measuring the spacing of assembly holes in large workpieces, this paper proposes a visual measurement scheme based on the collaborative measurement of multiple sets of binocular cameras. Four groups of binocular cameras are arranged above the measurement platform. To achieve the unification of the coordinate systems of each group of cameras, an exact correspondence relationship is established through the coding points in the common view area to eliminate mismatching, and the transformation matrix is solved by combining the least square method, which overcomes the defect of the traditional ICP point cloud registration algorithm being sensitive to the initial pose. The strategy of "image enhancement - sub-pixel detection - robust fitting" was proposed, and an adaptive threshold selection strategy based on the gray-scale distribution of the image was introduced, which optimized the screening process of edge points and improved the positioning accuracy of assembly holes. The measurement results show that within the working range of 3m × 3m, the measurement accuracy of the hole spacing is better than 0.1mm.