Using two D415 on the same FOV
Dear all
I was wondering how to improve the accuracy of the depth given by a D415 and these are the three options i was thinking:
1. try repeating the automatic calibration
2. take more than one scan and joining the results with a median filter (median over time, not over adjacent pixels)
3. scanning the same area with two D415 and matching the results.
In the latter case of course the two cameras cannot take the image at the same time. I think I read somewhere (but sorry I cannot find it anymore) that a special hardware is provided to sync and avoid interference of the structured light used by the two cameras to get the 3D depth info. Which is it and how it is supported by the SDK?
Thank you very much
Kind Regards
-
Hi Alessio Stella Structured Light is used by the RealSense SR300 and SR305 camera models. Multiple SR300 / SR305 cameras can interfere with each other.
The 400 Series cameras use a different depth sensing technology called Stereo. Multiple 400 Series cameras do not interfere with each other.
Using the Dynamic Calibrator tool or the On-Chip Calibration and the Tare Calibration that are built into the More side-panel menu of the RealSense Viewer can help to keep your camera's calibration healthy. As the On-Chip and Tare functions can be launched within the Viewer, they are useful for performing a quick depth health-check on your cameras.
https://dev.intelrealsense.com/docs/self-calibration-for-depth-cameras
There are a wide range of other ways to optimize the accuracy of your camera. An example of this is using the optimal depth accuracy resolution for your particular camera, which on D415 is 1280x720 (compared to 848x480 on D435 / D435i). Intel's camera tuning guide provides a broad range of advice for improving the results from the camera.
https://dev.intelrealsense.com/docs/tuning-depth-cameras-for-best-performance
You can overlap the FOV of more than one 400 Series camera. This can be used to increase the overall field of view (for example, four D435 cameras arranged in a horizontal arc for a 180 degree view).
https://www.intelrealsense.com/intel-realsense-volumetric-capture/
FOV overlapping can also improve depth data quality due to redundancy in the data from more than one camera observing the same area at the point where the views overlap. The more cameras that are in the arrangement, the less blind-spots that there will be in the observation of the scene.
Cameras can be hardware synced so that their timestamps try to follow a common master sync signal. This master signal can be transmitted from a "master" camera through physical sync cables attached to "slave" cameras, or with a wireless pulse from a signal generator.
It is not compulsory though to use hardware sync to capture views from multiple cameras though. As an example of this, the link below features an Intel guide to stitching together multiple real-time point clouds with more than one 400 Series camera in ROS.
https://github.com/IntelRealSense/realsense-ros/wiki/Showcase-of-using-2-cameras
If you do wish to explore the subject of multiple-camera hardware sync then the white-paper document in the link below should be helpful.
https://dev.intelrealsense.com/docs/multiple-depth-cameras-configuration
-
Thank you so much, sorry there was probably a misunderstanding. By structured light I meant here the dots pattern (i guess IR dots) that is projected on the image to improve the stereo depth extraction (that is what i had understood, please correct me if I am wrong). One can see this dots pattern also on the sample images of https://dev.intelrealsense.com/docs/self-calibration-for-depth-cameras
so i was wondering wether these dots might interfere between two D400 on the overlapped FOV
-
The semi-random dot pattern emitted by the built-in projector of the camera will not negatively affect the cameras when FOVs are overlapped. In fact, it can be beneficial to have an increased number of dots from multiple projectors present in the scene.
https://dev.intelrealsense.com/docs/projectors#section-4-increasing-range
It is worth bearing in mind though that the laser-based projector in the cameras can result in a noise effect on the image called laser speckle. This can be reduced by using an external LED based projector.
https://dev.intelrealsense.com/docs/projectors#section-7-led-projectors
Please sign in to leave a comment.
Comments
3 comments