3D reconstruction simulation method based on adaptive generalized orthogonal projection

With the advancement of machine vision, multi-vision measurement technology has found increasingly widespread applications in aviation, aerospace, military, and civil fields. To address the spatial intersection problem, we propose a generalized orthogonal projection simulation method to calculate a linear solution. Weight coefficients are determined based on the reprojection error of different image points, enabling weighted nonlinear optimization. For matching homonymous image points in low-texture or similar gray-scale distribution scenarios, we propose matching strategies based on the fundamental matrix, trifocal tensor, and the shortest distance between the reverse projection point and other reverse projection rays, depending on the number of effective cameras. Experimental results show that when noise intensity is 0.1 pixel, the RMSE of our algorithm is 0.34 mm, which is 26% lower than the DLT algorithm. At a measurement distance of 5000 mm, the RMSE is 0.24 mm, 17% lower than DLT. When measuring randomly distributed spatial points, our algorithm outperforms DLT in computational accuracy, robustness, and stability.