The 3D reconstruction method of a line-structured light vision sensor based on composite depth images

Abstract

Abstract In order to simplify the three-dimensional (3D) reconstruction process of a rotating scanning line-structured light vision sensor, a 3D reconstruction method of composite depth images is proposed. First, a frame of an image acquired by a camera at a constant speed is converted into a column of data in the composite pixel image. Second, the composite pixel image is transformed into a depth image via coordinate transformation. Finally, the 3D coordinates of the object are calculated from the depth image by perspective transformation. The parameters of the light plane are calibrated by using a one-dimensional target, and the focal length of the depth image is calibrated by using a ball target. Compared with the traditional 3D reconstruction method of a structured light vision sensor, this method uses a composite depth image instead of point cloud splicing, which is simple to operate and flexible to calibrate, and further improves the efficiency of the 3D reconstruction of a structured light vision sensor. Experiments show that the absolute error of the 3D measurement in the measurement range of 380–4000 mm is less than 0.21 mm, and the accuracy is higher than 0.26%, which significantly improves the measurement efficiency and point cloud density of the system, and can meet the general 3D reconstruction requirements of the scene.