Non-line-of-sight correction for target detection and identification in point clouds
Examples disclosed herein relate to an autonomous driving system in a vehicle having a radar system with a Non-Line-of-Sight (“NLOS”) correction module to correct for NLOS reflections prior to the radar system identifying targets in a path and a surrounding environment of the vehicle, and a sensor fusion module to receive information from the radar system on the identified targets and compare the information received from the radar system to information received from at least one sensor in the vehicle.
1. A sensor system, comprising:
a sensor module adapted to:
detect objects in a field of view of the sensor module;
detect reflections from a reflective surface outside the field of view;
generate a point cloud from reflections;
discretize angular space of the point cloud into a predetermined number of bins, and
correct for the reflective surface to detect objects in the field of view.
2. The sensor system as in claim 1 , wherein the sensor scans the field of view.
3. The sensor system as in claim 2 , wherein sensor module is a radar system having an array of radiating structures.
4. The sensor system as in claim 3 , wherein the sensor system is further configured to receive a point cloud from the radar system and generate a corrected point cloud.
5. The sensor system as in claim 1 , wherein the sensor system is part of an autonomous driving system, wherein the NLOS correction module receives a supplemental point cloud from the at least one sensor in the vehicle.
6. The sensor system as in claim 1 , wherein the reflective surface is a planar surface.
7. The sensor system of claim 1 , wherein the sensor module is further adapted to identify at least one planar reflecting surface using a supplemental point cloud.
8. The sensor system as in claim 1 , wherein the sensor module is further adapted to generate a point cloud of data representing sensor detections.
9. The sensor system as in claim 1 , wherein the sensor module is further adapted to generate corrected point cloud of data using detected reflections.
10. The sensor system as in claim 9 , wherein the sensor module is further adapted to map detected reflections.
11. The sensor system as in claim 1 , further comprising a perception module to determine a control action for the sensor module.
12. A method for operating a sensor system, the method comprising:
directing an antenna to generate RF beams at a plurality of directions in a field of view;
receiving reflected RF beams from targets in the field of view;
generating a point cloud from the reflected RF beams;
discretizing angular apace of the point cloud into a predetermined number of bins;
detecting objects from outside of the field of view from a reflective surface in the field of view; and
identifying at least one target from the corrected point cloud.
13. The method as in claim 12 , further comprising:
locating the reflective surface in the field of view.
14. The method as in claim 13 , further comprising applying a threshold to filter out spurious surfaces.
15. The method as in claim 14 , further comprising transforming the point cloud into spherical coordinate system.
16. The method as in claim 15 , further comprising extracting intersecting bins for each planar surface.
17. The method of claim 16 , further comprising extracting position and surface normal vector information for each planar surface.
18. The method as in claim 17 , further comprising define reflection operations defined by extracted position and the surface normal vector information.
19. The method as in claim 18 , further comprising remapping reflections to determine locations of targets.
20. The method as in claim 12 , further comprising extracting intersecting bins for each planar surface.