MartyG-RealSense
commented
2 years ago Hi @Cpeck71 With RealSense camera models that are in a casing and have front glass, the depth measurement is taken from the front glass and not the camera lenses behind the glass.
Adding -4.5 (mm) to the depth measurement to find the depth ground truth value at the lenses does not mean the depth will be a negative value, as it is subtracting 4.5 mm from the Z-depth measurement of the object or surface that is being observed by the camera.
In regard to from-front mechanical drawings of RealSense cameras it should be taken into consideration that from the camera user's perspective they will be behind the camera, and so the orientation of the imaging components will be reversed in the behind the camera perspective if the camera is rotated around so that the back of it is facing you.
The 400 Series stereo (left and right) depth cameras are an example of this, where the left imager is on the right -side of the camera on the frontal drawings.
Cpeck71
Before opening a new issue, we wanted to provide you with some useful suggestions (Click "Preview" above for a better view):
All users are welcomed to report bugs, ask questions, suggest or request enhancements and generally feel free to open new issue, even if they haven't followed any of the suggestions above :)
Camera Model | L515 -- | -- Firmware Version | 01.05.08.01 Operating System & Version | Linux - Ubuntu 18.04 Kernel Version (Linux Only) | 4.9.253-tegra Platform | Jetson Nano SDK Version | 2.49.0 Language | C++ Segment | Computer vision/Navigation
Issue Description
Hello RealSense Team,
I would like some clarification on the extrinsic parameters between the hardware components of the L515. For refrence, I've looked at issue #8249 and the linked extrinsics documentation and OpenCV pinhole camera model pages from that comment. From my own work, it is my understanding that the published data from the IMU agrees with the documentation and the IMU frame is oriented as the blue frame drawn in the figure below (I hope it's okay I'm using a screen grab from the datasheet).
To my issue: I run the rs-enumerate-devices tool with the '-c' option to get the calibrated extrinsic parameters between all of the hardware components (IMU, depth and camera). The trouble i have is that the extrinsic translation vector doesn't agree with the frames drawn here. The IMU and Depth camera are perfectly aligned, and the depth to camera (along with IMU to camera) rotation matrices are so close to identity that for the sake of this argument lets consider them to be nearly aligned such that all hardware frames have the same orientation.
Now, according to the rs-enumerate-devices the translation vector from Gyro to Depth is: T_G2D = [0.01245,-0.01642,-0.00057] from Gyro to Camera: T_G2C = [0.01214,-0.00253,-0.0053];
If you note, the X distance is positive, however according to the drawing the IMU should be to the left of the Camera and Depth frames, i.e. going from Gyro to depth/camera should be a negative translation. Furthermore, the Camera to Depth translation is: T_C2D = [0.000307,-0.013997,0.004785]
which doesn't make sense either since it should be positive in the Y and negative in the Z (since the origin of the depth frame is 4.5mm recessed from the glass cover) according to my current understanding of the orientation of the hardware.
Is there something I'm fundamentally missing from this or possibly an error in the enumerate devices tool?
Thanks so much for the help, Caleb