Project name: Radiometric Detection for Patrol Boats
Developer: Michael Civay
Description: Thermal solution that uses two stitched Lepton cameras to perform detection in harsh conditions.
About the project: Solution formed using Python and OpenCV running on a Raspberry Pi 3B+. Keypoint matching and homography matrix estimation was utilized to stitch frames from two FLIR camera video streams. This solution provides nearly 100 degrees of seamless detection with observed detection of target motion at night over 300 feet away. Establishing matched keypoints to form the homography matrix was challenging in dynamic conditions. Detecting logs or other debris saturated in the water with similar temperatures was difficult until they were a few feet from the cameras.
Future development will include SMS alerts, improved image recognition, tuning detection parameters and ruggedization of sensor mounting.
Here is a short video highlighting some of the use cases: