Research On Key Technology Of Point Cloud Surface Reconstruction For Workpiece Damage Detection

In recent years,with the continuous increase of railway transportation tasks and the increase of train running speed,people are paying more and more attention to its operational safety.In the long-term operation of the train,due to the fatigue wear of the wheels and rails,and the pressing of foreign objects,it is easy to cause the rail to break and cause major accidents such as rollover and derailment.Therefore,it is necessary and important to study the surface damage detection and visualization of railway wheel and rail workpiece.This article is mainly based on the three-dimensional reconstruction of the point cloud of the wheel and rail surface and the surface damage detection,The main research contents are as follows:Firstly,in order to achieve high-precision detection of railway wheel and rail workpieces,for the current stage,it is affected by the accuracy of the 3D scanner and the complex environment on site,it is difficult to use the 3D scanner to scan the wheel-rail workpieces on site.This paper proposes to independently build a CAD model of the wheel-rail workpiece and convert it into point cloud data.Aiming at the problem that the point cloud generated by the CAD model is difficult to maintain the sharp geometric features of the surface,this paper proposes a point cloud reduction algorithm based on curvature and bounding box.The algorithm divides the input point cloud into a series of voxel grids,and reduces the point cloud in the voxel grid according to the curvature characteristics.Experimental results show that the algorithm in this paper largely retains the sharp and edge feature points of the point cloud model,avoids the phenomenon of holes in flat areas,and provides a basis for subsequent efficient surface reconstruction and damage detection.Secondly,in order to improve the accuracy of the point cloud surface reconstruction algorithm,due to the influence of noise and outliers during the point cloud surface reconstruction process,the reconstructed surface is prone to artifacts,overfitting,holes,etc.This paper proposes a Poisson surface reconstruction algorithm based on Tikhonov regularization.The algorithm uses linear interpolation to interpolate sample points,and uses Tikhonov regularized L2 norm to smooth the original point cloud to reduce noise points and outliers,reduce the error caused by noise points and outliers.Through comparative experiments on the commonly used Standford 3D scanning library point cloud,Ein Scan-Pro scanning point cloud and the wheel and rail point cloud generated by the CAD model,it is proved that the reconstruction accuracy of the algorithm in this paper has been greatly improved,and it has certain feasibility and robustness.Finally,the surface damage of the wheel and rail point cloud is detected.First,building a surface damage on the point cloud model of the wheel and rail,and then using the ICP algorithm to register the damage data and the standard data,calculating the deviation between the two to judge the damage.Finally,the depth color mapping method is used to convert the damaged point cloud model into a color image,and the damage area is identified according to the color change on the color image.In order to visually display the damage area,the surface reconstruction algorithm proposed in this paper is used to visualize the damage data.