The UAV employs U-Net model for identification of forest fires

This paper explores a method for monitoring and identifying forest fires using drone remote sensing technology combined with the U-Net model and its improved attention mechanism. Traditional forest fire prevention measures are costly and have limited coverage, whereas drone technology offers an efficient and flexible solution for early detection of forest fires. In this study, the IEEE flame dataset was used to train and validate the improved U-Net model through high-resolution, multi-angle drone image data. Experimental results indicate that the U-Net model, augmented with channel attention and spatial attention mechanisms, performs excellently in complex backgrounds and high-resolution images, significantly enhancing the accuracy of fire area recognition and segmentation. The findings demonstrate that this method has notable advantages in identifying fire locations and accurately assessing fire conditions, providing robust technical support for the early warning and prevention of forest fires.