3D reconstruction and Visual Odometry
This project consists of 3 main parts :
-
Transformation and projection of a cloud of Lidar points on a camera (3D to 2D) : In this part you will use the transformations provided by the dataset to project a 3D point cloud (Lidar/Velodyne source) onto the image plane of the first color camera (left). This allows you to know the depth of a set of 2d points in the image, but also to know the color of the 3D points.
-
Create a 3D mesh from Lidar data points and camera images : Merge all 50 LIDAR acquisitions, after performing appropriate transformations, into a single point cloud, use the corresponding image data to color the point cloud. Save the results to a standard 3D file.
-
Visual odometry : Estimate the pose of the vihecule : Our objective is to calculate the trajectory traveled by the vehicle using a visual odometry approach. We will use the stereo image sequence provided by Kitti to calculate odometry. Next, we will compare the calculated trajectory with the data provided by the IMU.
