Research on train wheel point cloud registration algorithm based on key points by fusing Super-4PCS and ICP

Wheels are critical components of railway vehicles, and the dynamic measurement of wheel parameters is of paramount importance for the safe operation of trains.To enhance the matching accuracy in the existing dynamic measurement processes for train wheel parameters, this paper proposes an improved point cloud registration algorithm based on key point fusion of the Super Four-Points Congruent Sets (Super-4PCS) and Iterative Closest Point (ICP) algorithm. Firstly, point cloud filtering and normal estimation are performed on the wheel point cloud data to obtain source and target point clouds with normal information. Subsequently, the Intrinsic Shape Signatures (ISS) algorithm is employed to extract key points, and the Fast Point Feature Histograms (FPFH) point cloud feature descriptor is utilized to characterize the extracted key points. Then, a two-level registration strategy is used to improve registration accuracy, in which the Super-4PCS algorithm is applied for primary coarse registration and the ICP algorithm is used for the secondary fine registration, respectively. Finally, the experiment is conducted to validate the proposed algorithm and the performance of the algorithm is further comparative analyzed through the listed registration evaluation metrics. Experimental results demonstrate that the proposed algorithm significantly improves registration accuracy and robustness for wheel point cloud data, with the Root Mean Square Error (RMSE) reduced from 0.0631 to 0.0002, and the Mean Absolute Error (MAE) reduced from 0.0671 to 0.00026, compared to traditional algorithms. However, the algorithm's performance is sensitive to point cloud density and noise levels, and its effectiveness may vary under different environmental conditions.