cywputao / trajectory_processing

Trajectory Processing

This is the program that extracts trajectories from raw videos

Camera Calibration

This is the first step to read the parameters for Camera Matrix and Distortion Coefficiency, etc.

• Shoot a short video for a dashboard at different positions using a Static camera, and save it in the folder video
• Unfold the dashboard vedio using unfoldVideo.py, name the folder for saving the unfolded images accordingly
• Compute Camera Matrix and Distortion Coefficiency using CamCalibration.py

Remove Distortions in Images

This is the second step to remove the Distortion of the recorded videos

• Unfold the trajectory vedios using unfoldVideo.py, name the folder for saving the unfolded photos accordingly
• Remove the distortion using ImgUndistortion.py

Extracting Trajectories

This is the third step to extract the trajectories from the undistorted videos (images)

• Concatenate the undistorted images into video for Tracker http://physlets.org/tracker/ (manually tracking)
  • <ffmpeg -r 25 -i img%03d.png -c:v libx264 -vf fps=25 -pix_fmt yuv420p out.mp4>
• or Use SiameseFC tracker http://www.robots.ox.ac.uk/~luca/siamese-fc.html

Transform pixel coordinate system to real-world coordinate system

This is the 4th step to convert pixel trajectories to real-world trajectories

• Get four referencial points in both pixel coordinate and real-world coordinate systems respectively for affine transformation
• Run CoordinateTransfer.py

Trajectory Visualiztion

This is the last step to plot the trajectories

• Run plot_trajectory.py to visualize the extracted trajectories