Proceedings of the 7 th World Congress on Civil, Structural, and Environmental Engineering (CSEE'22) Lisbon, Portugal Virtual Conference – April 10 – 12, 2022 Paper No. ICEPTP 204 DOI: 10.11159/iceptp22.204 ICEPTP 204-1 Forest Fire Detection and Prediction from image processing using RCNN Abhay Chopde 1 , Ansh Magon 2 , Shreyas Bhatkar 3 1 Vishwakarma Institute of Technology 666, Upper Indira Nagar, Pune, India abhay.chopde@vit.edu 2 Vishwakarma Institute of Technology 666, Upper Indira Nagar, Pune, India ansh.magon19@vit.edu 3 Vishwakarma Institute of Technology 666, Upper Indira Nagar, Pune, India shreyas.bhatkar19@vit.edu Abstract- Forests are the most important part of terrestrial life as they provide shelter to 80% of terrestrial life and provide them with food as well. Forest fires are a serious threat to flora and fauna followed by deforestation. Various preventive measures are taken to avoid such incidents from taking place still, many such incidents happen every year causing long term damage to surrounding biology, environment, and wildlife. This paper proposes a large-scale monitoring system and deep learning-based forest fire detection model that can detect forest fires from video frames captured by UAV drones. The proposed CNN model successfully detects forest fires with 97.29% accuracy. This will help to control the forest fires before they get out of control. Keywords: Forest Fire prediction, RTH index, RCNN, UAV 1. Introduction Forests are the reason for most of the earth’s terrestrial biodiversity. Forest is an ecosystem that provides food, shelter to 80% of the living beings on the earth. Most of the time homo sapiens have existed, spent their lives in the forest. They play a crucial role in weather, atmosphere, rains and other various ecological factors important for the existence of terrestrial life. Forests are the largest terrestrial storehouse of carbon (and thus fuel). They are still home to many indigenous tribes. Around 60 million people are from various indigenous tribes around the globe. The area covered by forests has been shrinking due to deforestation. Forest fire is another threat to forests. Forest fires exist since the forests exist themselves. Forest fire is uncontrolled in vegetation more than 1.8 metres (6 feet) in tallness. These shoots frequently arrive at the extents of a significant blaze and are now and again started by ignition and hotness from surface and ground fires. A major woodland fire might crown-that is, spread quickly through the highest parts of the trees before including undergrowth or the timberland floor. Accordingly, fierce blow-ups are normal in wood flames, and they might expect the attributes of a firestorm. The natural cause of forest fires is mostly lightning or high atmospheric temperature and dryness. Human activities like spillage of crude oil or inflammable substances can also cause forest fires to start. Respective authorities implement various preventive measures to avoid forest fires. Tourists are also provided with instructions like selecting an open location for forest fire and extinguishing the fire completely before leaving. Still almost every year several incidents of forest fires take place. This indicates that prevention alone is not enough to avoid such incidents. Forest fires spread varies directly with factors like flammable materials available, speed of the wind, weather conditions etc. All these factors work in favour of fire, then in the very short period it covers a huge area. To avoid this paper proposes a deep-learning-based forest fire detection model that can detect the forest fire from satellite images.