A Cutting-Edge Approach to Forest Fire Region Identification Through Deep Learning

Our environment has been drastically impacted by global climate change. This global climate change is leading to many natural disasters which affect the livelihood of humans as well as animals. By burning timber coffers, a timber fire is a form of natural disaster that has a significant impact on people as well as creatures and shops that depend on the timber terrain. So, there is a need to detect these forest wildfires early and also the most occurring regions of wildfire. Therefore, the proposed system provides the deep learning model for forest wildfire region identification to limit the loss and damage caused by forest wildfires. The proposed method finds forest wildfire regions by using the image dataset where image frame categorization takes place. The system used the VGG19 network model, which classifies images as either fire or non-fire after being trained on a given quantity of image data. If the model predicted a fire image, the objects in the fire image were identified using the Yolov6 algorithm. Later by using temporal and spatial features, we detect the most occurring wildfire regions where we achieved an accuracy of 98.4%. The proposed system's objective is to identify areas where forest wildfires are likely to occur and notify at-risk communities in advance.