Every day in the summer here in California, we are exposed to extreme wildland firefighting incidents. To fight such wildfires, thousands of firefighters need to spread across hundreds of square miles. Recent years have witnessed an aggressive push towards building Unmanned Aerial Systems (UASs) for helping the fire-fighting missions. In this demo, we present a vision-based aerial sensing system for early fire detection with on-board intelligent processor. We propose a new open source perception system for joint autopiloting and multi-sensory object detection with a tight power budget. The technical approach focuses on developing a robust aerial sensing pipeline for fire detection in low-visible and smoky environments based on multi-cameras and thermal image sensor.
Available at: http://works.bepress.com/kaikai-liu/36/