Research On SFM 3D Reconstruction Technology Fusion Inertial Navigation Data

With the continuous development of information technology,traditional two-dimensional image information can no longer fully meet the needs of social development,and people’s demand for three-dimensional information is becoming more and more urgent.In recent years,methods of using images to recover the three-dimensional structure of objects in the scene have been widely studied.Among them,the Structure From Motion(SFM)method is the current research hotspot.It efficiently realizes the three-dimensional reconstruction of the target scene through the information-rich,low-cost images.However,due to interfering factors such as occlusion and rapid movement of the camera image,the visual relevance is weak,which often affects the efficiency of the 3D reconstruction work.The characteristics of high sampling frequency and reliable instantaneous accuracy of the Inertial Measurement Unit(IMU)are used to provide external posture information for the photos taken by the camera to make up for the weakness of weak visual association.This paper is based on the SFM 3D reconstruction method which integrates the mobile inertial navigation data.The main research contents are as follows:(1)In order to obtain the effective parameters for the operation of the SFM 3D reconstruction system,this paper studies the camera model,camera calibration,IMU calibration,camera and IMU joint calibration,image feature extraction and detection methods and other related issues.This kind of data evaluates the reliability of the traditional SFM 3D reconstruction system.(2)Aiming at the problem that the SFM 3D reconstruction system is highly dependent on the input image and time complexity,the inertial navigation information is introduced,the Kalman filter method is used to solve the problem of IMU data noise,and the influence of accelerometer bias and gravity is eliminated.The IMU integral model is established,and the IMU data is introduced into the traditional reprojection error optimization equation as the constraint term of image transformation matrix.The IMU data is used as the initial iteration value of the error optimization algorithm to optimize the camera pose and point cloud position,so as to reduce the number of iterations,improve the reconstruction results,and improve the efficiency and accuracy of the system.The experimental results show that compared with the traditional SFM method,the system error is reduced by 35%,the number of generated point clouds is increased by 20%,and the running time is reduced by 14%.(3)Aiming at the point cloud hole repairing problem in the 3D reconstruction process,this paper studies the point cloud hole repairing strategy based on the SFM method,which mainly includes key technologies such as point cloud filtering,density processing,hole boundary recognition,point cloud registration,and related experiments.The experimental results show that using the point cloud hole repair technology,more complete point cloud results can be obtained,and the root mean square error of three groups of point cloud hole repair experiments is less than 0 ± The surface area difference is less than 0.5%.