SPRACX2 April   2021 TDA4VM , TDA4VM-Q1

 

  1.   Trademarks
  2. 1Introduction
  3. 2What is Structure From Motion?
  4. 3Introduction to Occupancy Grid Mapping
  5. 4From Point-Cloud to OG Map
  6. 5Algorithm Flow: SFM-Based OG Mapping
  7. 6Algorithm Flow: SFM-Based OG Mapping on TDA4VM
  8. 7First Example Implementation on TDA4VM
  9. 8Second Example implementation on TDA4VM
  10. 9References

4 From Point-Cloud to OG Map

To create a world-centric OG Map, first each point in the point-cloud generated by the SFM algorithm, is placed in a world-centric coordinate system using vehicle location data from the Inertial Navigation System, or INS, and sensor calibration data. Sensor calibration data includes the relative positions of the different sensors with respect to each other. Then, each point is classified as either originating from the ground (ground point) or from an object (object point). Next, the state of each cell in the OG map is determined using the world-centric point-cloud as follows -- if a predetermined number of object points are found in a cell, it is deemed occupied; if a predetermined number of ground points are found in a cell, it is deemed free; if neither criterion is met it is deemed unknown.