Abstract:Obtaining the 3D surface and depth information of the workpiece target is the key to realizing industrial stereo vision applications. A method of combining two sets of binocular vision systems was proposed. The randomly placed workpieces were acquired in multiple directions and the three-dimensional surface point cloud of the target workpiece was obtained. Among them,the two sets of binocular vision systems would generate two sets of parallax images of the working scene according to the NCC (Normalized Cross Correlation) matching algorithm,and correct the stereo depth information after segmentation and denoising. In the process,a novel conversion method was adopted,and parallax was adopted. The stereoscopic depth information of the pixel points of each coordinate position in the image in the x,y,and z directions was converted into the gray value of the pixel at the corresponding position in the X,Y,and Z images,respectively. The two sets of stereo depth data were sampled and stored together in the calibration reference camera coordinate system for information fusion. Finally,the 3D reconstruction experiment of the randomly placed industrial workpieces can restore the clear contour point cloud for parts with different overlap,height and posture,and also produce a clear level in the overlapping area.