Get current location from existing pointcloud or data.
I have very accurate pointcloud data that I get from Terrestrial scanners. Is there a method of using existing data or CAD models to get a robots current location in reference to the existing data? I am looking for a way to do this programmatically so that it can be done in real time in 5 seconds or less. If it is possible to do a point set registration with the current camera data.
Thanks MartyG. I have an existing pointcloud generated from November 2019. I have referenced the pointcloud to the longitude and latitude of the earth (and to XYZ positions). I want to send the robot back into this area and use the existing pointcloud or reference points to determine the robots positions. This should reduce drift from the tracking camera. I will know (ballpark) the initial starting point of the robot. Is there a compare method that will give the current scene position to a reference scene(or cloud)
There is an open source tool that can be used for comparing point clouds that is called CloudCompare.
Possibly OPEN3D -
Point Cloud Library (PCL) can also do comparisons apparently, so that may be a better choice if you want to integrate your project with the RealSense SDK as the SDK has a support wrapper for PCL. The RealSense SDK does not have a support wrapper for Open3D yet.
Google 'compare point cloud pcl' for more details. Here's an example:
The 400 Series cameras have a default sensing range of 10 meters. In theory, maximum distance (MaxZ) increases if the Disparity Shift value is decreased instead of increased. As error increases linearly with distance though, the measured results would probably be too inaccurate to make it worth extending sensing range that far.
I believe the forthcoming new L515 lidar camera does not have this decrease of accuracy over distance, and it also never has to be recalibrated during its lifetime. It is suited to use indoors or under controlled lighting conditions though and has a maximum range of 9 meters.
Please sign in to leave a comment.