A method for 3D reconstruction of train carriages based on LiDAR point clouds

Abstract

In the context of the continuous growth of railway transportation volume, the safe operation and maintenance of train carriages have placed higher demands on the accuracy and efficiency of damage detection. Traditional detection methods are unable to effectively quantify various types of damage such as corrosion, dents, and surface deformations on the carriages. The digital detection technology based on 3D reconstruction provides a new approach to this problem. However, most existing 3D reconstruction methods rely on multi-view registration, which has the drawbacks of complex processes and long time consumption, making it difficult to meet the actual requirements of rapid detection during train operation. Therefore, this paper proposes a lightweight single-station laser radar reconstruction method without multiview registration. The comparative experiments conducted on four different types of carriages have verified that the comprehensive performance of this method is excellent. The BPA algorithm is prone to surface overfitting and loss of key features, while the Alpha-shape algorithm is prone to generating small holes, disordered structures, and poor noise resistance. The minimum Chamfer distance (CD) of this method can reach 0.006346 millimeters, the point-to-grid distance (PTD) is stable between 0.098 and 0.227 millimeters, and the reconstruction time is controlled within 5.34 to 6.38 seconds. This method demonstrates excellent feasibility and applicability in 3D reconstruction of train carriages, and can provide an efficient and feasible technical solution for the digital detection and intelligent maintenance of train carriages.