Among the current stereo matching algorithms for moving targets in substations, there are a few matching points and mismatching problems. This paper proposes a three-dimensional matching algorithm for moving targets in smart substations that combines the A-KAZE algorithm and the improved Speeded Up Robust Features (SURF) algorithm. The A-KAZE algorithm is used to extract the matching feature points of two images, and second-order multi-scale heating is added to the SURF feature vector to further calculate the secondary response. A high threshold algorithm is used to increase the matching points, and the random sampling consensus algorithm eliminates the mismatch points to complete the matching. The effectiveness of the algorithm is verified through experimental comparison. The experimental results show that, compared with the previous matching point pairs from 908 to 1202 pairs, the matching accuracy is increased from 92.51% to 96.17%, which has definite practical value. This work is supported by the Science and Technology Project of Fujian Province (No. 52130A170006).