- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
I'm using the D435 camera for obstacle avoidance on a drone. The drone has 3 cameras mounted. An I use launch the RS without pointcloud but launch a separate depth_imag_proc node to get an xyz pointcloud from the cameras.
I attached the launch file so you see what parameters are used. Also I set the visual preset to HIGH_ACCURACY for all cameras.
Now the problem is, that objects/spots appear in the pointcloud where there are none. Those spots do not look like the average noise to me, as they are really small point clusters and not just random noise. One thing they have in common is, that they are a temporal phenomenon. They only exist in one pointcloud and are gone in the next. I attached a video from a test flight. You can ignore all the colored markers, important for this case is the visualization of the pointcloud. We are flying around a single tree at a height of about 4 meters. There are no other obstacles present and all the smaller point clusters to the right of the drone correspond to the mentioned phenomenon.
Do you have any idea where that is coming from? Do you think I need to change any of the parameters?
Any help is highly appreciated
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
It occurs to me that as it is an outdoor setting, it may just be bouncing off flying insects that are traveling so fast that they only register for an instant as the camera pings off them. They are therefore not in the next point cloud because they really have gone!
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Thank you for your fast reply!
But I am not so convinced, the bigger spots in the video are definitely too large for insects. Also the drone causes quite some air movement, soI don't think there are any insects that close. Do you have any other idea to what could be causing them? Because the seriously limit the performance of our algorithm and we need to get rid of them.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Given how bright the sunlight is in the recording, it may be worth reviewing the outdoor sunlight section of Intel's camera tuning guidelines. Left-click on the image below to view in full size.
Noteworthy is the advice that in strong light, "exposure can be reduced to near 1ms, which reduces motion artifacts as well"
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page