Research On Fast Three-dimensional Flaw Detection And Recognition Method Of Reflective Surface

In the recent years,shining objects surface three-dimensional defect detection has been a hotspot and difficulty in the field of 3D measurement.Owning to the shining reflection,cameras will be saturated,resulting in the existence of holes and noises in the generated three-dimensional point cloud,which will directly affect the accuracy of the three-dimensional defect detection.What's more,most of the traditional three-dimensional defect identification is to analyze the generated point cloud,or to analyze the topological connection structure between the points cloud,causing a large load on the computer,so that the computational efficiency is low.In this dissertation,the two mentioned problems above has been studied thoroughly and corresponding solutions are put forward.Firstly,a set of binocular structured light measurement system based on three-frequency six-step phase shift and heterodyne multi-frequency is built and calibrated for the three-dimensional defect detection of reflective objects.This dissertation has studied the problem converted from camera saturation caused by reflection of object surface into the insufficient camera dynamic range,introducing the multiple exposure fusion technique to strengthen the dynamic range of imaging effect.This method could stop the object light reflection largely and retain the image details without distortion to provide an important guarantee to the subsequent three-dimensional defect inspection compared with traditional light reflection technique such as spraying developing agent,retrofitting polarization system to camera and image pixel fitting etc.Secondly,for the large calculation load caused by the huge point clouds,a three-dimensional defect detection method combining two-dimension and three-dimension is put forward in the dissertation.This method first fits a standard plane out of the three-dimensional measuring point cloud and conducts fitting alignment of the fitting plane and point cloud plane to calculate the distance from points deviating the standard plane to the standard plane.Then mapping is conducted between the distance and color gradation image to locate the defect of three-dimensional point clouds via the color differences of two-dimensional levels image.After the location,color decoding is made on the coordinate position to get the specific dimension of the defect.This method raised in the dissertation has lowered the calculated amount traversed by point clouds to some extent,shortened the traversing duration of three-dimensional point clouds and improved the efficiency of three-dimensional defect detection.Finally,the three-dimensional measuring system set up in the dissertation has conducted a three-dimensional measurement and defect detection of the surface of metal cartridge prone to reflect light.Concave pits defects have been identified by our flaw detection method,and the results are compared with those of Spectroscopic confocal apparatus.The experimental results show that the method proposed in this dissertation is less than 0.05mm,and the validity of the method is proved.