I have an issue with my astras: I’m capturing a point cloud and converting it to world space to match a scene that contains Vive headset and controller objects, so the point cloud from the astra should fit – i.e. hands captured in the astra should appear where the controllers are. This works fine when using a kinect, but the astra is always off. We’ve tried manually calibrating, but if you get one hand to line up with a controller, the other is always positioned some distance away from its controller.
My entire pipeline is to grab the astra point cloud, then translate and rotate that to world space (no scaling), there’s nothing else going on. Looking at the calibration utility I noticed it doesn’t even generate distortion values, so I’m guessing that the images are not being undistorted? Or is that built into the camera, and happening before I get the data?
Not undistorting the image (or a not-great factory calibration of the intrinsics) is my only guess as to why there’d be a difference between the kinect and astra, but if anyone knows of another reason I’d love to hear it.