Satellite pose estimation

Goals

No deep learning based approachs.

Playing around with some basic image processing functions in order to estimate the 6-dof pose of a satellite (known model) using a monocular calibrated camera.

The code is in C++ and based on OpenCV functions.

Results

First tests

	Canny	Poses
Dark background 1	Canny_Dark_background_1_video4.mp4	Poses_Dark_background_1_video4.mp4
Dark background 2	Canny_Dark_background_2_video6.mp4	Poses_Dark_background_2_video6.mp4
Earth background 1	Canny_Earth_background_1_video2.mp4	Poses_Earth_background_1_video2.mp4
Earth background 2	Canny_Earth_background_2_video5.mp4	Poses_Earth_background_2_video5.mp4

Second tests: another "cost function" + illumination correction

	Old vs New
Dark background 1	Poses_Dark_background_3_video4.mp4
Dark background 2	Poses_Dark_background_4_video6.mp4
Earth background 1	Poses_Earth_background_3_video2.mp4
Earth background 2	Poses_Earth_background_4_video5.mp4

Third tests: old vs fix merge line segments

	Old vs New
Dark background 1	Poses_Dark_background_5_video4.mp4
Dark background 2	Poses_Dark_background_6_video6.mp4
Earth background 1	Poses_Earth_background_5_video2.mp4
Earth background 2	Poses_Earth_background_6_video5.mp4