Research And Development Of 3D Reconstruction System For Painting Of Civil Aircraft Body Parts

The field of 3D vision has always been the core component of computer vision,and its application in the industrial field is particularly important and critical.In industrial production,the use of three-dimensional visual recognition technology can greatly improve production efficiency and industrial intelligence.This paper applies the 3D reconstruction technology in the field of 3D vision to the industrial production line of civil aircraft sheet metal painting,and performs 3D reconstruction of aircraft sheet metal parts in the production line workshop to obtain 3D models for identification,and According to the spatial positioning of the part,design the spraying trajectory to complete the spraying automation work.Aiming at the image acquisition of parts,a four-eye stereo vision system is designed in this paper to collect images,so as to acquire images containing complete part information for input to the system and reconstruction.In addition,two different image segmentation methods are designed based on the degree of color difference between the background and parts of the scene.That is,when the color of the background and the part is large,the watershed algorithm is used for segmentation.When the color difference is small,the Grab Cut algorithm is used.Divide.The advantage of this solution is that on the premise of ensuring the quality of the segmentation,two methods are used to segment the scene parts in different situations as quickly as possible,and to prepare for the subsequent model reconstruction.In the reconstruction module,Sift algorithm is used to extract feature points.The data structure of kd-tree is used to calculate nearest neighbor matching.BA workflow optimization is performed at each step to obtain sparse point cloud and SFM algorithm.Dense point cloud under PMVS algorithm.In order to reduce the reconstruction speed caused by the difference in the number of parts on the stage,a different solution is proposed in this paper.When the number of parts on a stage is less than 6,even the image segmentation is performed first,and then threedimensional.Reconstruction,so that each part is modeled separately;when the number of parts is greater than or equal to 6,the overall reconstruction of the stage is used,and finally the parts in the 3D model are divided to obtain each Model of the part.This scheme is a method to maximize reconstruction efficiency while ensuring reconstruction quality.In terms of surface reconstruction,this paper uses the conventional Poisson(Poisson)surface reconstruction algorithm to complete.When it comes to the surface extraction in the algorithm flow,the surface extraction algorithm based on Marching Cube is used,and the obvious noise in the model is removed by the reconstructed mesh area to reduce the signal-to-noise ratio.Finally,based on the reason that the images contain different vertices,two texture rendering methods are proposed to fully recover the 3D information of the part.After the algorithm research is completed,this paper presents an engineering design scheme based on the actual scenario to fit the industrial assembly line operations.And completed the software system.