A Study On True Color 3D Detection Technology Based On Binocular Stereo Vision And Structured Light Scanning

Three-dimensional digital technology is an important research content in the field of intelligent manufacturing.Because of the fast and non-contact characteristic of three-dimensional optical scanning,it is widely used in 3D measurement,reverse engineering and heritage preservation.Therefore,how to reconstruct the real object model quickly and accurately is the key to the research.In this paper,based on binocular structure light scanning true color 3D detection technology,making full use of the advantages of binocular vision sensor non-contact measurement and color three-channel calibration principle based on light intensity,the coordinate data and color information of object surface are obtained quickly,and the reconstruction of object true color model and the measurement of 3D topography are finally realized.In this paper,based on the research of common 3D measurement technology and system,aiming at the problem of scanning blind area in general line structured light sensor scanning system due to camera angle and object surface occlusion,a binocular structured light detection principle is proposed and the detection model is established.By analyzing the conversion relationship between the planar and line-structured light planes of dual cameras,the scanning model is simplified,and the system calibration and accuracy verification are carried out.Secondly,according to the irregular shape of the object surface and the different reflection condition,the sub-pixel precision of the line structure Light strip center is extracted by using the gradient centroid method.Aiming at the shortcoming that the true color information of the surface can not be obtained by ordinary line structured light detection,the color information conversion and correction of the surface is realized according to the color three-channel calibration principle based on light intensity,and the true color information of the surface is obtained.Thirdly,aiming at the problem of 3D point cloud registration of binocular structured light system,the 2D image coordinates of the object obtained by the two cameras are converted into 3D world coordinates by the calibration result of binocular structured light system,and the 3D color point cloud is formed by fusing the color information of the corresponding points.After the pretreatment of denoising and thinning,two color point clouds are registrated together by ICP algorithm based on k-d tree search.The 3D model of the object is reconstructed by splicing results,and the surface topography parameters are obtained.Finally,the model is used to scan the vehicle model.The result shows that the accuracy of the calibration method is up to 0.02 mm,the sub-pixel center of structure light strip can be quickly extracted by gradient centroid method,the precision of scanning system is improved,the result of point cloud registration and color reconstruction is obtained,the correctness and validity of the system is verified.