boylesr
commented
10 years ago The main hurdle with this requirement is ensuring the same human is not counted twice, one for both the Kinect and thermal sensor. One solution is to use some type of distance algorithm to determine if there are one or two humans detected.
If a map is generated and fed back to the user, the coordinates for potential humans should be separated into three cases: 1) Detected by Kinect but not IR (likely dead). 2) Detected by IR but not Kinect (likely covered) 3) Detected by both.
This could serve as a rough level of triage.
Would also have to take into account the possibility of counting two humans as one. Devil's advocate being two people together being recorded as one, One individual leaves for help and collapses later on.
Description: When a human is detected by either the Kinect or thermal sensor, the human detection system needs to communicate with the mapping/tracking system so that when a heat signature or skeleton is located, that point can be correlated to location on map. Since the Kinect has a depth sensor, if the UAV knows its height the human's exact coordinates may be determined with the depth value and the tilt angle of the Kinect. If the thermal sensor locates a potential human but the Kinect does not, the UAV's current position can be recorded as the sensor does not have depth data.
Dependencies: Navigation
Fit Criterion: Multiple test cases - 1) Have an uncovered human be detected and have it's coordinates returned/recorded. 2) Have mostly covered human detected by thermal sensor but not Kinect, and have the UAV's current position recorded.
Supporting Materials: N/A