-
Notifications
You must be signed in to change notification settings - Fork 4.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
L515 camera "box_dimension_calculate_error" #8929
Comments
@MartyG-RealSense @RealSenseSupport Hi, In the "convert_depth_frame_to_pointcloud" method, I multiplied the depth image by 0.00025 instead of 0.001. But how is the "depth_thresold" value specified in the "calculate_cumulative_pointcloud" method determined? Can you help me? (The camera I work with is L515.) |
@MartyG-RealSense |
Hi @zeynepkoyun Our main box dimensioning software is the DWS software/SDK product for L515 is here: https://www.intelrealsense.com/dimensional-weight-software/ Regarding the python example, not all of our examples have been updated properly to support the L515 camera. We'll need to look into this further. |
Thanks for your answer @RealSenseSupport I started to investigate, but I want your opinion on a topic. Could the distortion pattern defined as None [0 0 0 0]] in the depth image for the L515 camera have an effect on the calculation? When do you intend to update your Python instance? |
Hi @zeynepkoyun |
@maloel Thank you for your answer. When I visualized the PointCloud data, I saw a lot of scatter in the product at the corners of the object, so I couldn't find the exact box size without clearing the noise data. |
Hi @zeynepkoyun, |
I didn't get any errors when I worked with the D415 camera. I also took advantage of the source I mentioned above and this source ("https://greenvalleytl.com/wp-content/lidar360_en/ToolReference/DataManagement/PointCloudTools/OutlierRemoval.html"). |
There was work done on the example, here is a link to the PR: #9188 |
Hi @zeynepkoyun Do you require further assistance with this case, please? Thanks! |
@MartyG-RealSense |
Thanks very much @zeynepkoyun for the update - I'm pleased to hear that you found a solution! |
Hello @zeynepkoyun , Can you share what solved the problem for you? We discovered that, after correct calibration, the "world coordinates" of the checkerboard are correct for x-y, but z is wrong. We expect that the checkerboard plane should be z=0 , because the depth_threshold of -0.01 would then keep all the points that are more than 1 cm above the checkerboard, i.e. the box, but not the board itself. However, we find that the points of our checkerboard have average z=2.4 m. There are no points <-0.01, and therefore the cumulative_pointcloud is empty. On further inspection, we find that the checkerboard plane at z = 2.4 m is roughly 2 times the distance between camera and plane, d = 1.2 m. This seems to suggest a "+" instead of "-" solution: z = d + d = 2.4 instead of z = d - d = 0. Alternatively, a solution could be in what you "negative errors" due to scattering of photons in LiDAR data. So please, can you share what solved the problem for you? |
By changing helper_functions.py line 180 into: z = depth_image.flatten()/4000 , we get "reasonable" results for single L515 camera setup (the bounding box is too large). We expect that multicam setup will not work for L515 cameras in default asynchronous mode: both cameras will emit IR pulses at the same time, thus interfering each other's depth readings. There is a whitepaper on using multiple L515 with a hardware trigger/sync. cable which should work for static object/scenery measurements. We are looking for a L515 multicam setup for dynamic object/scenery. Any suggestions welcome. |
Hello, I was able to run the code "box_dimensioner_multicam_demo.py" for the D415 camera. However, the L515 cannot find the objects for the camera. When I looked at the code, I noticed that the pointcloud value was empty. What should the deep_thresold value be for this camera?
The text was updated successfully, but these errors were encountered: