Georeferenced Mosaics for Tracking Fires Using Unmanned Miniature Air Vehicles

Fire tracking is an increasingly important area of research aiming to help firefighters more effectively fight fires. In the past, piloted aircraft have been the main source for obtaining fire characteristics from the air. In recent years unmanned air vehicles (UAVs) have become popular for the same purpose. While large UAVs are an effective means of assisting firefighters, they are expensive to purchase and operate. Mini UAVs or MAVs, on the other hand, have become cheap and reliable platforms for surveillance missions. However, due to weight and size constraints, MAVs are often equipped with error-prone sensors, in contrast to large UAVs, resulting in poor-quality georeferencing and geolocating. Techniques to solve this problem in the fire tracking context have focused on fusing information from on-board infrared and color cameras and using various filters to correct single geolocation estimates. However, a fire can be very large and arbitrarily shaped, thereby invalidating the single geolocation point assumption. We aim to solve this problem by producing a Georeferenced Uncertainty Mosaic (GUM) in which size, shape, and geolocation information is shown simultaneously in an easy to understand georeferenced image. The GUM is created by appropriately blurring the infrared images captured by the on-board camera, and using the blurred images as observations in a particle filter.