Research On Robust 3D Reconstruction Method For Weakly Textured Objects

With the rapid development of machine vision technology,3D reconstruction technology has become an important method for measuring the shape of objects in industry.For 3D reconstruction of weakly textured objects,it is difficult to match enough feature points with traditional methods,resulting in incomplete reconstruction.Therefore,in order to meet the requirements of high accuracy in visual measurement,this paper proposes a comprehensive 3D reconstruction method that takes into account the subject and details of the reconstructed object,and reconstructs the surface and details of weakly textured objects with high accuracy.First,explore the principle of the 3D reconstruction method,analyze the camera calibration model and feature matching principle model in the traditional incremental 3D reconstruction method flow,study the traditional reconstruction methods of linear triangulation and nonlinear beam adjustment,and analyze weakly textured objects Based on the difficulties of traditional reconstruction methods,a preliminary 3D reconstruction process was established.Then,a point cloud densification model based on patch expansion filtering reconstruction algorithm is established,and the point cloud density is enriched by spatial patch expansion.An analytic hierarchy process is designed to screen for mismatched feature points,and the number of feature points is controlled using a pre-trained neural network model.Then,the reconstruction method based on depth map fusion is studied,and the depth information is used to supplement and reconstruct the above-mentioned unreconstructed details to obtain a more complete point cloud model.Based on this,aiming at the problem of holes in the depth map under high light environment,a lighting compensation method for light parameter estimation and a depth map smoothing scheme based on mean shift algorithm are designed.The standard test pictures were used to simulate the feasibility of this method.Finally,the above three-dimensional reconstruction method is experimentally verified.First,a reconstruction experiment is performed on a weakly textured object in an indoor scene.After obtaining a dense point cloud model,the point cloud is expanded into a triangular mesh model,and the model gap is optimized using Poisson algorithm Then,the model surface is rendered to obtain a complete 3D reconstruction model of the weakly textured object.On this basis,the experimental verification of the weakly textured objects in the outdoor environment is performed again,and the experimental data is evaluated using various performance indicators.The results show that the accuracy of the reconstructed model surface parameters can be controlled within 1 ? 2mm,and the model structure is similar.At about 90%,the performance indicators meet the requirements of visual measurement,which verifies the robustness and effectiveness of the method in this paper.Based on the effect of 3D reconstruction and the practicability of the algorithm,this paper proposes a set of 3D reconstruction solutions for the surface of weakly textured objects.The research content provides theoretical and technical support for industrial visual measurement.