Research On 3D Measurement And Defect Extraction For High-speed Railway Train Body Surface

The high-speed rail is an important card for China’s rapid rise and a comprehensive reflection of international influence and status.The train body is the most important large-scale complex structure on the high-speed rail.Putty grinding,which determines the quality and efficiency of the production,is an important process of the production of high-speed train body.The current main method is manual grinding,which is pretty much time-consuming,environmentally Polluting and has low quality stability.Applying industrial robot to putty grinding can improve processing efficiency and quality stability.As the putty is applied manually,randomly distributed bumpy defects in size and location are generated on putty surface.Bumpy defects areas and flat areas requires different processing parameters,but it’s difficult to change processing parameters at uncertain positions by the currently used robot processing methods.Visual detection was applied in this paper to distinguish between defect areas and flat areas.Therefore,processing parameters can be changed according to the detection results.Issues related to the calibration and data acquisition of 3D visual detection systems,point cloud data preprocessing and feature extraction were studied in this thesis.The main contents and innovations of this article are as follows:(1)According to the characteristics of high-speed train body,a two-degree-of-freedom line structure light measurement system was designed.The CCD camera,laser plane,and the scan direction of the detection system were calibrated through chessboard.Standard gauge blocks were used to verify the calibration accuracy.(2)Point cloud data collected by the detection system was preprocessed.To be specific,statistical analysis and bilateral filtering were applied to denoise and smooth point cloud;a modified angle-chord deviation method was used to simplify point cloud;due to the lack of certain features on the surface of the train body,a cloud splicing method based on visual calibration was proposed.The splicing algorithm was verified through the experiment on standard saw-tooth block.(3)As the train body doesn’t have strict theoretical basis,A method of defect extraction based on self-fitting benchmarks was proposed.Firstly,a random sampling consistency segmentation algorithm based on the maximum principal curvature is proposed,dividing the point cloud into the planar area point cloud and the surface area point cloud.Then the benchmarks were fitted in each area.An improved doublethreshold method of normal vector angle and curvature local feature weight value was proposed to identify sharp defects on the putty surface and a projection distance threshold method based on fitting datum was proposed to identify gradually fluctuating defects.The experiment of surface defects detection was conducted at the high-speed train body grinding experiment platform,and the validity of the defects detection algorithm was proved by the result.At last,a comparative grinding experiment of the visual guided processing method and the off-line programming method was performed on high-speed train body sample to verify the purposed visual guided processing method in this paper.