Detecting wildfires using unmanned aerial vehicle with near infrared optical imaging sensor

Abstract

The severity of wildland vegetation fires is expected to grow in response to climate change. Therefore, the price of combating fires will likewise go up while still posing a serious risk to the firefighters. Various countries have invested enormous sums of money in combating fires throughout the years, and this trend is expected to continue. This provides compelling reasons for surveillance systems that can track and detect wildfires at early stages. The Optronic Sensor Systems of the Council for Scientific and Industrial Research (CSIR) in South Africa is developing a small, cost-effective Near-Infrared (NIR) optical imaging payload for tactical forest fire-fighting operations. This paper reports on the field measurement from sensors that detect NIR spectral emissions from the electronically exited alkali metal Potassium (K) emitted during the flaming phase of the biomass. The NIR sensor consists of a combination of two optical imaging systems (target and reference sensor) placed side-by-side with common (identical) field of view. The concept uses images obtained from the optical imaging systems are compared to determine the pixels which are much brighter in the target band relative to the reference band, which are defined as fire detections. This principle uses a portable imaging system consisting of two similar complementary metal oxide semiconductor (CMOS) cameras with high sensitivity within the NIR band. The fire detection is computed using the K-line ratio algorithm. The results presented in this paper show that it is possible to perform early fire detection of biomass fires using low cost NIR sensors coupled with advanced image processing algorithms.