Method and system for multiple 3D sensor calibration
A method for multiple 3D sensor calibration for a passenger conveyance system, the process including a computing a centroid location from spatial measurements for each of a pair of 3D sensors that form a common field of view via a moving object mathematical model; computing translation parameters from the locations of each centroid location; and correcting one of the pair of 3D sensors to a common world coordinate system from the translation parameters.
1. A method for multiple 3D sensor calibration without taking measurements in sensor and world coordinate systems, the process comprising:
calculating a centroid location from spatial measurements for each of a pair of 3D sensors that form a common field of view via a moving object mathematical model;
calculating at least one of translation and rotation parameters from the locations of each centroid location;
synchronizing each of the 3D sensors in time; and
correcting one of the pair of 3D sensors to a common world coordinate system from at least one of the translation and rotation parameters.
2. The method as recited in claim 1 , wherein individually calibrating each of a pair of multiple 3D sensors to world coordinates results in zero relative rotation there between.
3. The method as recited in claim 1 , further comprising synchronizing each of the multiple 3D sensors in time comprises synchronizing to within 1 frame time.
4. The method as recited in claim 1 , further comprising calculating from the (x, y) locations of the two centroids (x 1 , y1) and (x 2 , y2), at least one translation parameter by subtraction.
5. The method as recited in claim 1 , further comprising calculating from the (x, y) locations of the two centroids (x 1 , y1) and (x 2 , y2), at least one rotation parameter.
6. The method as recited in claim 1 , further comprising calculating from the (x, y) locations of at least two centroids (x 1 ,y 1 ) and (x 2 ,y 2 ), the translation parameters (dx, dy) by a Least Squares (LS) solution.
7. The method as recited in claim 1 , wherein one of the pair of 3D sensors is chosen as governing with respect to the common world coordinate system.
8. The method as recited in claim 1 , further comprising individually calibrating each of a multiple 3D sensors to world coordinates.
9. The method as recited in claim 1 , wherein three measurements are performed of a common point in response to three translation parameters being unknown.
10. The method as recited in claim 1 , further comprising calculating the translation (dz) by finding individual ground planes of each sensor and computing a difference.
11. The method as recited in claim 10 , further comprising using a random sample consensus (RANSAC) algorithm to find a consistent set of points to find each of the individual ground planes.
12. The method as recited in claim 1 , wherein six measurements are performed of a common point in response to three rotation and three translation parameters being unknown.
13. The method as recited in claim 12 , further comprising using a top and a bottom of a moving object and a separation between the top and the bottom of the moving object when the to three rotation and three translation parameters are unknown.
14. The method as recited in claim 1 , wherein an object mathematical model is employed to estimate the common centroid location.
15. The method as recited in claim 14 , wherein the object mathematical model is an elliptical cross-section mathematical model.
16. The method as recited in claim 14 , wherein the object mathematical model is an elliptical cross-section person mathematical model.
17. The method as recited in claim 14 , wherein the object mathematical model is a full 3D person shape mathematical model.
18. The method as recited in claim 1 , wherein synchronizing each of the multiple 3D sensors in time comprises synchronizing to within 1 depth map sample time.
19. The method as recited in claim 1 , wherein synchronizing each of the multiple 3D sensors in time comprises synchronizing to within 1/25 th of a second.
20. The method as recited in claim 1 , wherein synchronizing each of the multiple 3D sensors in time comprises synchronizing to within a depth map sample time small enough that moving objects are approximately stationary during the sample time.
21. A sensor calibration system for a passenger conveyance system without taking measurements in both sensor and world coordinate systems, comprising:
a multiple of sensors with at least one overlapping field of view;
a processing module in communication with the multiple sensors, the processing module including a moving object mathematical model operable to estimate a common centroid location to compute translation parameters from the locations of each centroid location to facilitate correction of at least one of the pair of 3D sensors to a common world coordinate system from the translation parameters, each of the multiple of 3D sensors synchronized in time.
22. The system as recited in claim 21 , wherein the object mathematical model is an elliptical cross-section mathematical model.
23. The system as recited in claim 21 , wherein the object mathematical model is a full 3D person shape mathematical model.
24. The system as recited in claim 21 , wherein each of the multiple of sensors generate a depth map in 1D, 2D, 3D, or a combination thereof.