Monocular Visual Odometry for Tello Drone

Author: Muye Jia

The pipeline starts with camera calibration for the RGB camera located on the front of the drone; then the images captured will be converted to grayscale, and subsequently fed to ORB or ShiTomasi feature extractor. Features extracted from the previous frame will be matched in the next frame to find the image coordinates of the same set of features in the previous frame and current frame, which can then be used to solve the epipolar constraint equation for the essential matrix that contains the camera rotation and translation matrix representing the pose transformation between the two frames. Next, the transformation matrices will be stitched together to form the entire camera trajectory

Results

Real-time visual odometry on DJI Tello Drone

draw_points.mp4

The following are the trajectories obtained from real deployment on DJI Tello Drone:

Camera Calibration

1. Calibration photos using a grided chessboard need to be taken using the target camera, from at least 10 different angles and views. Run python3 cam_calibrate.py with the take_calibration_photos function enabled, and the host computer connect to the DJI Tello drone WiFi; the script will take photos using Tello drone camera at at a frequency of 10Hz.

2. Next, use the photos taken and run python3 cam_calibrate.py with the calibrate_camera function enabled, the cam_matrix variable contains the intrinsic parameters of the camera.

Visual Odometry Test

1. Using KITTI dataset as validation. Run python3 visual_odometry.py with plot_KITTI function enabled to test the VO pipeline on the KITTI dataset, one can choose to enable the bundle-adjustment.

2. To test the VO on the drone, the user need to record images using the drone and then perform the odometry offline by running python3 visual_odometry.py with plot_drone function enabled.