high speed capture for D415
Hi,
I am scanning and reconstruting object using D415 in high speeds. The depth quality turns out to has low accuracy. I'm trying to analyze the reason but few information found. I understand D435 is better suit for speed, but obviously its accuracy is not for 3D scan.
1. is there a speed limit about D415 for 3D scanning? if exceeds, then would not get the depth image accurately.
2. is there any instruction about using a certain exposure time at a certain speed?
3. any test paper about D415 depth quality on speed?
thank you.
-
Hi Jianfei Li The D415 has a rolling shutter on its depth and RGB sensors, whilst the D435 model has a rolling shutter on its RGB sensor but a fast global shutter for the depth sensor.
If motion of the D415 camera or an observed object moving in front of the camera exceeds a certain speed then the image can blur. A good example of the speed of motion that may cause blur with D415 is the quick waving of a hand in front of the camera, as demonstrated in the D415 discussion at the link below.
https://github.com/IntelRealSense/librealsense/issues/3554
I would recommend compensating for the slow shutter speed by setting the frames per second (FPS) speed to 60 or 90 FPS at 848x480 depth resolution if possible.
If auto-exposure is disabled and a manual exposure time is used then lag may be introduced if using an exposure value that is within a certain range, as described here:
https://github.com/IntelRealSense/librealsense/issues/1957#issuecomment-400715579
RealSense supports a High Speed Capture mode for camera models with a global shutter on their depth sensor such as D435. Intel's high speed capture white-paper guide at the link below discusses rolling shutter (D415) versus global shutter (D435).
In regards to depth accuracy, there can be a range of reasons why depth values have inaccuracy. This could include the strength of lighting in the scene, the type of lighting, the color of the observed object, whether the object has a plain / smooth or highly detailed surface texture, the distance of the object from the camera and how reflective the object is. If you can provide an RGB image of the observed scene then that will be useful for diagnosing possible causes of depth error.
Please sign in to leave a comment.
Comments
1 comment