Research On 3d Reconstruction Method For Highly Reflective Surface

With the development of robotic automatic picking,self-assembly and industrial detection technology,the industry needs more and more high-precision 3D information for object recognition and positioning,surface detection,quality control and so on.3D reconstruction plays a very important role in robotics.Even in common tasks such as robotic automatic picking and sorting and industrial detection,it has been recognized that it is necessary to use 3D measurement to achieve functions that are not possible with existing 2D measurement.Academia and industry have studied many solutions for high-precision 3D reconstruction.Since there are many objects with highly reflective surface in practical application,how to reconstruct the surface of highly reflective objects has become a very challenging problem.In the existing structured light 3D reconstruction method,there is a problem that the point cloud data is easily lost due to the object-reflection,which leads to the hole of point cloud data,this paper proposes a 3D reconstruction method of the surface of highly reflective object based on the fusion of multi-angle monocular structured light system.In this method,the two cameras of the system share the same projector,respectively forming two sets of monocular structured light systems.Then,the predetermined rigid transformation matrix is used to improve the fusion quality of the point cloud fusion region.Finally,a point cloud fusion decision map is generated to avoid repeated reconstruction of the fusion region of the left system and the right system,and ensures the accuracy of the non-fusion region to achieve the accuracy of monocular structured light system.Aiming at the problems that the calibration method of the existing structured light system is not accurate and difficult to operate,a calibration method of structured light system based on Zhang Zhengyou calibration method and local homography principle is proposed.In this calibration method,the projector is treated as an inverse camera.Through the transformation relationship established between the camera and the projector,the corner points in the calibration plate image acquired by the camera are mapped to the coordinates of the projector image by using the local homography principle,and the calibration of the projector is performed by a method similar to the camera calibration.Then,the calibration of the monocular structured light 3D reconstruction system is converted into a calibration of the binocular camera system.At the same time,based on the in-depth study of the existing structured light coding method,an orthogonal gray code coding method is designed to make the system more robust and faster in 3D reconstruction of high reflective surface.Moreover,this coding pattern can be used for system calibration as well as for 3D reconstruction of the system,which reduces the complexity of the system design.We designed several experiments to verify the validity and accuracy of the proposed method.In addition,3D reconstruction experiments of complex objects such as bright metal surface,the middle frame of mobile phone and motor stator were carried out respectively.The experimental results show that the proposed method has the ability to solve the 3D reconstruction problem of highly reflective objects and occlusion problem of complex objects.