Research On Key Technologies Of 3D Imaging Of Textureless Target Surface Based On Binocular Structured Light

As an important branch of computer vision technology,three-dimensional vision imaging technology has become a research hotspot at home and abroad in recent years because of its high efficiency,high accuracy and ability to meet various requirements of automatic production line.Among them,structured light three-dimensional imaging technology has been widely used because of its low cost,high accuracy,high real-time,easy implementation and non-contact advantages.According to the bionic three-dimensional imaging model,two cameras are used to capture the measured object from left and right angles,the triangulation model is formed with the measured object to calculate the depth information of the target surface and obtain the three-dimensional image of the object.Because there are fewer feature points on the surface of textureless three-dimensional objects,this paper adopts the method of adding feature points artificially,projecting the color structured light encoded by De Bruijn cyclic sequence to the surface of the object as feature information,taking the fringe center line as feature measurement point,matching the left and right image feature points,and calculating the depth of the textureless three-dimensional object surface with calibration data,obtain three-dimensional point clouds.Among them,the main work of this paper is embodied in the following aspects:1.The hardware platform and software implementation of the binocular structured light-based three-dimensional imaging system are completed,and an easy-to-operate integrated interface is developed.2.The coding of 5-element 3-order De Bruijn color structured light with smaller fringe pixel width is proposed to increase the number of target measurement points in the image and make the reconstructed point cloud denser.Aiming at the problem that the Euclidean distance of adjacent fringes is too small and the interference is too large,a method of projecting structured light of different colors in the coded image separately is proposed.An improved coding method based on complementary codes is proposed and the corresponding decoding algorithm is designed,which further reduces the interference between adjacent pixels at the boundary and improves the robustness of the system.3.In order to improve the reconstruction accuracy,this paper uses the method of sub-pixel edge line extraction to determine fringe edge,and combines roughlylocating position to filter out noise points to extract sub-pixel center line.The left and right fringe sequence matching method based on Smith Waterman dynamic programming algorithm is improved to solve the problem of codewords missing due to occlusion destroying the uniqueness of sequence window,and to improve the accuracy and robustness of matching.Finally,aiming at the problem of sparse point clouds,we use depth completion technology to complete part of the empty pixels in depth map,and then reverse the method to obtain the depth information of the empty pixels to complete the densification of point clouds.4.Through three-dimensional reconstruction of different models and accuracy analysis,it shows that the system proposed in this paper can restore the three-dimensional contour of real object surface with high accuracy.