Using Thermal Imaging to Efficiently Detect Wildfires from a UAV

by Amay Shah

A few words from the participant(s)

What steps did you take to develop your project?

This project uses a thermal imaging camera, interfaced with Arduino, and connects it to a drone to accurately and efficiently detect wildfires. Fabrication and testing occurred in two parts: Beta I and Beta II. 

     The beta I test, after software development but prior to attaching to the camera to the drone, measured how the angle of the camera in relation to the fire affected the time taken for the program to output an alert. From a set distance, the camera had 100% accuracy in detecting the fire of a candle. The final beta II testing evaluated the performance of the camera attached to the drone in a real environment, using a campfire to simulate a forest fire, while varying distance and measuring the time taken till fire alert. Although increasing distances of up to 90 feet led to slower times for the fire alert in the program, all fires were accurately detected within 15 seconds, meeting the initial engineering goal of the project. The data also showed a strong linear correlation between time for wildfire alert and distance with an r^2 value of 0.941 and an ANOVA test p-value under 0.05. The outcome of the beta I and II tests reveal that the thermal imaging camera, using the developed fire recognition program, is capable of detecting wildfires in a timely and accurate manner, ready to be utilized in real-world scenarios.

Why are you competing?

Regardless of distance and angle, the final product of the engineering design process, a UAV with a thermal imaging camera using my program interfaced through Arduino,  can detect wildfires in a timely and accurate manner, meeting the initial engineering goal of the project. 

     Multiple phases of testing demonstrate that the camera and program are successful and capable of detecting wildfires in a relatively fast manner. Since the drone will not always be directly above the fire, the angle test proves the accuracy at 45 degrees is significant.      Furthermore, the abundance of fauna in many forests requires the camera to not associate the temperatures of wildlife with that of a forest fire, another successful test. Lastly, the camera’s capability of working on a drone flying 90 feet high proves that it can be used in forest situations. In the future, drones with higher flying capabilities may be used to test this software at higher altitudes. 

     The product can be used in forests all around the world to provide a more cost-effective method of fire detection than satellite imagery and more efficient than human patrols. Areas without access to satellite surveillance can use this product to detect wildfires and prevent widespread destruction.