Computer vision for wildfire detection: a critical brief review

In this critical brief review, we explore the pivotal role of computer vision in wildfire detection, following the PRISMA methodology and focusing on 35 key studies published between 2018 and 2023. Notably, convolutional neural networks, including models like YOLOv5, Inception v3, MobileNetV2, and Faster R-CNN, have emerged as the preferred choice for researchers in this field. Object detection emerges as the predominant computer vision task employed for wildfire identification. The review underscores a rising trend where researchers opt to utilize existing image datasets or create their own, incorporating various imaging modalities, from conventional RGB to thermal and infrared imagery. Unmanned aerial vehicles have gained increasing prominence for data collection, though they come with challenges such as limited battery life and data transmission bottlenecks. While alternative deployment methods like ground stations are considered, the review reveals a significant gap in literature regarding the practical deployment of satellite systems and advance monitoring systems for wildfire detection, pointing to a need for comprehensive studies on their operational viability and maintenance costs. Overall, this study aims to broaden the understanding of the complex interplay between wildfire detection and computer vision, highlighting the need for future solutions to be both technologically innovative and operationally viable.