Research on Composite Rapid Non-contact Detection Method for Surface Defects of Cylindrical Roller

Abstract:
In the production process of cylindrical rollers, due to the influence of technology and equipment, surface defects are easily generated. The surface defects of the rollers will cause the bearing to break and fail due to vibration during operation, so the detection of surface defects is particularly important. In order to quickly and accurately identify defects, this paper proposes a composite fast non-contact defect detection method based on 2D images and 3D point cloud data. First, image processing technologies such as OTSU threshold segmentation, Blob analysis combined with difference method are used to accurately locate all suspected defects and reduce the missed detection rate. Then use the α shapes algorithm to quickly and effectively extract the data boundary points projected to the two-dimensional plane, and then select the least squares method and B-spline interpolation to perform curve fitting to obtain defect depth information and reduce the false detection rate, and finally complete the research of the entire detection method. After experimental verification, the surface defect areas of cylindrical rollers with diameters of 1.5mm, 2.0mm, and 3.0mm are 12, 17, and 49 pixels, respectively, and the minimum depths of central scratches are 0.0681mm, 0.0656mm, and 0.1071mm, respectively. The whole method has important industrial application value.