Hi,
I try to use COLMAP to get the depth map of image sequence (video frames) taken by UAV camera, while no intrinsic parameters are known. The scene is crowded people near some buildings and it is generally far away from the camera.
I first try automatic reconstruction with command like:
/src/exe/automatic_reconstructor \
--workspace_path $PROJECT_PATH \
--image_path $PROJECT_PATH/images \
-- single_camera 1
but the performance is bad, so I try to put more control and do it step by step like use sequential_matcher rather than exhaustive_matcher since the sequence is from continuous frames in a video. But the performance is also not good.
Therefore, I want to know is there any thing I should do in order to get good performance with such long distance, small motion , continuous video sequence?
Thank you for your help!