Application of image classification in forest fire detection using UAV

This paper focuses on the application of image classification in forest fire detection using unmanned aerial vehicles (UAVs), discussing the development history of UAV image classification and the significance of machine vision in fire monitoring. Initially, the dataset used for fire detection and the data processing and enhancement techniques are introduced. Subsequently, the construction and architecture of the image classification model are detailed. The core of this study is to enhance the accuracy of model image recognition in complex forest environments by replacing optimizers, modifying the model architecture, and adding modules. Various models and optimizers are compared and analyzed, and the operations and significance of enhancement methods and attention mechanisms are explored. The aim is to improve training effectiveness through these strategies, thereby effectively supporting UAVs in forest fire detection.