what kind of environment is best for the pushbroom stereo algorithm

[sherry1222]

Hi, your algorithm is very innovative and has a great influence to my research.
However, i find that the pushbroom stereo algorithm can not detect some body parts, such as legs or head, while a person as an obstacle. So i really want to know what specific environment is best for this algorithm.
Is there a standard for obstacle selection? Can you tell me how to choose them to make me continue my research?
Thanks you for your help!

[andybarry]

Pushbroom stereo should work fine on body parts. The underlying stereo algorithm is very standard and should work in any textured environment. I tested it indoors at various distances using footage from people walking towards and away from the camera.

Are you sure you have your calibration set up correctly?

[sherry1222]

Thank you for answering my question!
I directly use the rectified images as the input to detect matching regions. So i think my results have no related to calibration. I saw an edge map which an image uses a Laplacian, the loss body part is lack of edges information. I guess it may not distinguish with the background.
About horizontal self-similarity, detections on horizon have good results. However, even though the detections eliminate most of wrong matching blocks, there still extremely few wrong blocks that cannot be fully removed. This question very puzzles me. Did this situation happen to you before?
Thanks again for your time!

[vishav7982]

Sir, I just Want to implement pushbroom stereo in python as my college project .
Can you tell me which file in your project is doing so ?
Thanks in advance !

Seeing your code will help me better understand it ! so that i can implement it in python .