3D Point Cloud Registration Method for Substation Robot Patrol Tracks

In cases where sensors cannot satisfy relevant prescribed conditions, the point cloud data composing the inspection track of a substation robot cannot be accurately matched. Therefore, a three-dimensional point cloud registration method based on infrared image feature fusion is proposed for the inspection track of a substation robot. The gradient histogram of the robot motion direction and local self-similarity description are extracted, that is, the HOG and LSS features. Both types of features are fused using a multi-feature adaptive fusion method. The key points of the fused trajectory features and optimal target trajectory pose parameters are obtained through a preliminary registration of the three-dimensional point cloud. The optimized iterative nearest-point algorithm is used to accurately register the patrol trajectory and improve the registration results of the patrol trajectory pose. The experimental results show that the feature fusion effect of the proposed method is satisfactory and can improve the edge clarity of the image. The deviation index after fusion is less than 0.2, and the registration of key points for different image sizes is accurately completed. Moreover, the inspection track after the registration is consistent with the expected track.