3D model reconstruction from a single image with IMU data

This research developed a method for reconstructing 3D scenes from a single photograph with depth map and inertial measurement unit (IMU). The proposed method uses IMU to provide the relative attitudes of Exterior Orientation Parameters (EOP) and assumes the camera position is the origin. Depth map provides the relative distance from feature points to the camera center. Therefore, collinearity condition equations can be constructed to determine 3D positions of feature points from a single image. In this study, there are four steps for 3D scene reconstruction from a single image: (1) camera and boresight calibration; (2) depth map generation; (3) feature points calculation; (4) 3D scene model reconstruction. The proposed method begins with angles calibration between camera boresight and IMU system. Broesight angles can be predicted using the result of image back projection and simultaneously recorded orientation data from the IMU. Using the relative distance provided from the depth map, coordinates of the extracted feature points are determined from collinearity condition equations. Finally, the feature points are used to reconstruct 3D models. Textures are then applied to the constructed models for visualization. In order to evaluate the quality of reconstructed model, this research simulates the effects caused by different errors to estimate the accuracy of the models.