Research On Surface Defect Detection Method Of Solid Workpiece Based On Differential Geometry

The detection of the surface defects of products in industrial production can effectively control the quality of industrial products and is an indispensable step in intelligent manufacturing.The use of defect detection based on machine vision can effectively improve the efficiency,stability and accuracy of inspection,thereby improving the quality of products?production efficiency,and effectively enhancing the market competitiveness of enterprises.In recent years,defect detection algorithms based on machine vision have developed rapidly,and they have gradually shifted from processing on two-dimensional planes to three-dimensional stereo vision to obtain detailed defect parameters.However,most of the current high-accuracy 3D data acquisition methods require high hardware,such as laser triangulation and raster projection.At present,the method of acquiring three-dimensional data based on binocular vision can effectively reduce the hardware requirements,but the depth information obtained through the stereo matching algorithm is only accurate in certain specific scenarios.Based on this,this paper proposes a stereo vision-based defect detection method,introduces the knowledge of surface theory in differential geometry,and obtains the required depth information by establishing a surface parameter model to realize the surface defect detection of conical parts.The thesis focuses on the research of defect detection technology for conical workpieces with low surface texture.For high-precision color images collected by binocular cameras,the contour curves of conical products and defects are extracted after image segmentation and surface parameters are established accordingly The model finally obtains the length and area information of the defect on the cone surface.(1)Analyze the image characteristics of the collected workpiece images in different color spaces,perform image segmentation in the HSV color space,and extract the contour curves of the workpiece to be tested and its defective parts.(2)Perform experimental statistics and analysis on the matching accuracy of many common matching algorithms on the data set,combined with the shape characteristics of the workpiece itself,propose a depth information acquisition method based on differential geometry,to avoid the huge calculation amount when the parallax range is large,At the same time,the accuracy of the acquired depth information is improved.(3)By analyzing the geometric relationship between the geometric feature parameters of the conical surface in three-dimensional space and the number of pixels on the edge of the product in the planar image,the quantity conversion formula is directly established.Therefore,only a small amount of feature point data is needed to recover the spatial parameter model of the conical surface and to obtain the depth information of the defect based on this.(4)By analyzing the relationship between the arc length and area of the closed curve on the curved surface and the first basic quantity of the curved surface,the geometric characteristic data of the defective part on the tapered product is obtained.The simulation results show that the error between the geometric quantity of the defect part obtained by this method and the real defect data is small,and the accuracy is greatly improved compared to the three-dimensional data obtained directly by the matching algorithm.