Once we identify human signatures in the data collected, we need to accurately report their GPS locations to enable in time and successful rescue. Following graphic illustrates our signature GPS location estimation algorithm:

 

 

Apr 5, 2017

We reviewed the algorithm to ensure its correctness and added the functionality to directly read the logs generated by the signature detection algorithms containing information about timestamps of the images in which signatures were identified, corresponding pixel locations of signatures in the images and the types of signatures. We did another round of testing but later found that the camera recording start time was not in sync with the actual time and thus the results were incorrect.

 

Mar 22, 2017

We implemented the first version of the algorithm in Python. We conducted the first round of test tests to test the algorithm but did not get good results because of some GPS data logging issues. There were some intervals of time for which there was no GPS data resulting in poor performance of the algorithm.