Research On Key Techniques Of Automatic 3D Reconstruction Based On Viewpoint Planning

Phase-assisted 3D measurement is widely used in many fields due to its fast measurement speed and high precision.With the development of application requirements and technology,the development trend of 3D digital display automation,how to automatically and quickly obtain 3D models is the focus of research in recent years.When performing three-dimensional measurement,since the sensor has limited field of view and needs multiple measurements,all the three-dimensional data can be obtained.At present,it is inevitable to rely on manual operation to bring repeated measurement and missing measurement.In order to make a one-time measurement or automatic measurement,viewpoint planning has important significance.Therefore,this paper focuses on the key technologies of viewpoint planning in automatic 3D reconstruction.The main contents are summarized as follows:The imaging constraints of the 3D sensor during the 3D reconstruction process are described,and the measurement space size of the system is estimated by establishing a mathematical model.According to the known object model,a viewpoint planning method is proposed.The main idea is that after the three-dimensional sensor satisfies the corresponding constraints,the number of unknown point clouds is covered at most,and the coverage of the entire model is completed with as few viewpoints as possible.It provides guidance for one-time measurement of the entire object or automatic measurement,and reduces labor intensity and improves work efficiency.A viewpoint planning method is proposed for the unknown model.The lowprecision digitized model of the object to be tested is quickly obtained by the depth camera.Then,based on the viewpoint planning algorithm,the point cloud data in the low-precision digitized model is visually planned,and the three-dimensional sensor is calculated respectively.Multiple viewpoint positions and line of sight directions when three-dimensional scanning of the object to be scanned is required.This can effectively improve the accuracy and scientificity of the sensor pose determination,and greatlyimprove the efficiency of the viewpoint planning,reduce the time consumption of the entire three-dimensional measurement process,and improve work efficiency.At the same time,the path optimization of the post-viewpoint scanning is studied.The shortest path is found by the genetic algorithm,which reduces the moving distance during the sensor scanning process and improves the efficiency.In order to realize the automation of 3D reconstruction,an efficient and flexible measurement method was designed,and a 3D measurement system based on robot arm and rotary table was built.After the calibration of the entire system sensor is completed,the viewpoint viewpoint and the line of sight direction of the object to be tested are determined by the proposed viewpoint planning algorithm,and the multi-viewpoint of the single three-dimensional sensor is realized under the control of the robot arm and the rotary table.,fully automatic 3D data acquisition and measurement.