Calibration Of The 3D Reconstruction System Based On A Line Laser

With the development of image processing and measurement technology,the application of 3D reconstruction technology has become more and more widespread,and the ways of 3D reconstruction are also more diverse.Different applications often use different reconstruction methods.The method in this paper uses a line laser to project onto the surface of the object to form laser stripes and combines with a motion platform to complete the slicing and reorganization of the reconstructed object.This paper optimizes the center extraction method of laser stripes,focus on the system calibration method,and finally realize the 3D reconstruction of small objects.The center extraction of laser stripes is a key step in the calibration and 3D reconstruction of the system.The laser stripes not only provide the feature points on the object surface,but also the plane where the laser is located provides an additional constraint to solve the problem that the image point and object point are not one-to-one in monocular reconstruction.After the extraction of the laser region,two mainstream center extraction algorithms are introduced in the paper,the gray centroid method and the steger algorithm.The second method is improved by using dynamic convolution kernels to adapt the laser stripes of different backgrounds and optimizing the problem of multiple center points when the stripes are too thick,ensuring the continuity and uniqueness of the center points.System calibration mainly includes camera calibration,line laser plane calibration and moving platform calibration.Camera calibration adopts the widely used Zhang Zhengyou calibration method to obtain the internal and external parameters of the camera.Line laser plane calibration is to find out the parameters of the plane where the laser is located,and three calibration methods are introduced in the paper.The method based on cross ratio invariance is improved,increasing the number of fitted points to improve the robustness and accuracy of the results.And the direct fitting method is also proposed,using all the extracted stripe centroids as fitted points of the light plane.Thousands of fitted points are available in just two pictures.And a new evaluation method is proposed to reflect the quality of the line laser plane calibration and the goodness of the center extraction algorithm in an integrated way,which uses a single image to measure the height of different objects before reconstruction.For the height measurement,the direct resolution method is used instead of the traditional triangulation method,which does not require the laser plane to be strictly perpendicular to the carrier platform or to obtain the corresponding angle.The calibration of moving platform is to get the motion vector of the platform,and then realize the slicing and reorganization in the scanning process.The reconstruction process will be simpler when the motion direction is parallel to the y-axis in the world coordinate system.Applying the above improved methods,the 3D reconstruction of small objects is achieved in this paper.