Research On 3D Reconstruction Based On Micro Inertial Sensors

3D reconstruction is becoming one of the most hot research in the field of computer vision, especially, it is important significant to reconstruct for the large-scale real-time scene. With the rapid development and wide application of micro inertial sensor, it is common in all kinds of intelligent devices, so the research and application of micro inertial sensor have been widely attention.This paper is researching 3D reconstruction against this background. On the basis of a large number of reference literature, the key technology of 3D reconstruction was studid, the main work of this paper is as follows:First, according to the camera calibration, this paper analyzed the principles and the applications of the Zhengyou Zhang camera calibration method and the camera calibration method based on Kruppa equations and applications. Then the paper introduced a camera calibration method based on micro inertial sensors, including the related knowledge about the micro inertial sensors used in the method, how to acquire the sensors’ datas when the images were taken and how to use the sensors’ datas to solve the camera parameters. The results of calculation through the algorithm were applied to the 3D reconstruction.Second, according to feature extraction and matching, this paper mainly analyzed Harris, SIFT and SURF, by comparing the several kinds of algorithm and the requirement of feature extraction and matching, this paper finally selected the SIFT algorithm as the method of feature extraction and matching, and through the RANSAC algorithm to eliminate false matching points in order to get accurate and robust matching point, which laid good foundation for 3D reconstruction.Third, in view of the 3D reconstruction of scene structure, this paper mainly studid the production of the 3D point cloud, the 3D coordinate was calculated using the triangle method, then the paper proposed the reprojection for the 3D point coordinates to generate the scene 3D point cloud model.Finally, the paper designed a complete 3D reconstruction system, the system contains the acquisition of the images and the sensors’ data, camera calibration, feature matching and 3D point cloud. Due to the increased use of micro inertial sensor, camera calibration was easy for the camera model estimation and calculation. And the accuracy of camera calibration is increased. Compared with traditional methods, this algorithm has advantages for low cost, high precision and simple operation. The system will be wide application in the navigation system, the remote sensing system and the industrial automation system, so much as entertainment, film and television.