Fire detection using infrared images for UAV-based forest fire surveillance

TLDR

UAV‑based computer vision systems are increasingly employed for forest fire surveillance and detection. This study proposes an automatic infrared image processing method for detecting forest fires from UAV footage. The algorithm combines histogram‑based segmentation of hot objects with optical flow motion analysis, followed by morphological filtering and blob counting to isolate and track fire pixels. Experiments demonstrate that the method reliably extracts and tracks fire pixels in infrared video sequences.

Abstract

Unmanned aerial vehicle (UAV) based computer vision system, as a more and more promising option for forest fires surveillance and detection, is now widely employed. In this paper, an image processing method for the application to UAV is presented for the automatic detection of forest fires in infrared (IR) images. The presented algorithm makes use of brightness and motion clues along with image processing techniques based on histogram-based segmentation and optical flow approach for fire pixels detection. First, the histogram-based segmentation is used to extract the hot objects as fire candidate regions. Then, the optical flow method is adopted to calculate motion vectors of the candidate regions. The motion vectors are also further analyzed to distinguish fires from other fire analogues. Through performing morphological operations and blob counter method, a fire can be finally tracked in each IR image. Experimental results verified that the designed method can effectively extract and track fire pixels in IR video sequences.