In view of the waste of computing resources and the lower positioning accuracy with the matching pixel points substituted into positioning model in binocular vision system，a 3D reconstruction algorithm based on the minimum of Euclidean distance is presented. According to the geometrical structure of nonparallel binocular vision positioning，a calculation model from the middle point of the common perpendicular line in different planes is established by solving the minimum value of two-point Euclidean distance. And then the effectiveness of the algorithm is verified by the experiments. Finally， the distance between the pairs of markers is measured by the calibrated system，and the error is less than 1 mm when the distance is 1.5 m from the binocular vision sensing units. The experimental results show that the algorithm has greater reconstruction accuracy and better real-time performance than the traditional method. At the same time，the method offers excellent robustness.