Research On 3D Measurement Technology Of Surface Defects Of Curved Parts Based On Multi-View Images

As the composite energetic material,polymer bonded explosive(PBX)have extensive applications in the national defense industry.The surface defects on it during the manufacture process will severely affect the property,and the difference in the position and size of the surface defect will have different effects on the product.Therefore,it has important practical value to develope a three-dimensional measurement technology for surface defects.In the field of industrial inspection,the image-based defect detection method can effectively obtain the two-dimensional contour of the defect,but cannot obtain the three-dimensional information.By combining the 3D reconstruction technology,the mapping relationship between the two-dimensional contour of the defect and the three-dimensional space can be established to realize the 3D measurement of the defect.Aiming at the demand for three-dimensional measurement of surface defects of PBX products,this paper uses two image-based defect 3D measurement schemes,and carries out related technical research,and designs related experiments to verify the feasibility and effectiveness of the scheme.The main contents are as follows:1.Combining with the automation PBX booster production and the simple surface structure characteristics,and given the measurement flexibility and cost,a 3D reconstruction method based on monocular camera and surface contour constraints of bulk products is proposed.Determine the pose of the camera according to the marked points pasted on the surface of objects.Then,an intersection equation of camera model and the surface contour can be established and the mapping of the pixel to 3D space can be realized.2.For the measurement requirements of small amounts of PBX products with complex surface or unknown structural parameters,a 3D measurement method for defects combined with binocular grating projection reconstruction and texture mapping is adopted.Use the camera calibration and rotatory axis calibration to establish the mapping relationship between point cloud and image of products.And a fast occlusion judgment algorithm baed on distance constraints is designed when the point cloud is projected to the image.Finally,a weighted color fusion method is used to generate the realistic color point cloud,and an accurate positioning method based on four-quadrant neighborhood search and linear interpolation is introduced to realize the reconstruction of defect pixels.3.The experiment systems are established to verify the 3D reconstruction result and defect measurement accuracy.Aiming at the reconstruction of known simple products and using the monocular measurement system including robots and industrial color camera,the reconstructions for products with and without marker points are conducted and the position and area of surface defects are measured.For products with unknown topography,through a defect 3D measurement system including grating reconstruction equipment and a color camera,the realistic model generation and defect measurement experiments have been carried out,and the defect size,location and area have been measured.