Hi,
I am trying to compute the time to contact (ttc) for moving (non-rigid) objects using moving monocular camera videos by computing the change of object size in each frame.
theoretically, the method is OK. practically, I got unstable results (e.g. increasing and decreasing numbers while the object is approaching).
I tried to reduce the temporal noise between every two consecutive frames by calculating the ttc over a time window (for example 10 frames). this reduced the inaccuracy but still no clear pattern in the results. I would expect the ttc to be increasing when the object is moving away and to be decreasing when the object is approaching, but I can't get such a result!
it is worth to mention that both the camera and the object are moving.
could anyone help me in this?
did I miss something?
Thanks
Did you try to link the output of the cameras? and keep the size as cue?
@Aparna Sathya Murthy, thanks for your reply.
I am using an off-the-shelf camera with normal lens and FOV. it is challenging process since the moving object keeps changing its direction of motion and speed all the time. in addition, the camera itself is moving. so it is hard to have a stable pattern for the ttc. the results are fluctuating significantly and that's what makes it hard to extract an accurate time to collision/contact.
Thanks
@Vera Maura Fernandes de Lima, I'm afraid that I didn't get your idea! what do you mean by linking the output of the cameras? I am using single monocular camera.
Thanks
- Badges
- Science topic
- Similar topics
- Computer Communications (Networks)
- Simulators