FCDM: An Improved Forest Fire Classification and Detection Model Based on YOLOv5

Intense, large-scale forest fires are damaging and very challenging to control. Locations, where various types of fire behavior occur, vary depending on environmental factors. According the burning site degree damage, this paper considers classification identification surface canopy fires. Deep learning-based detection uses convolutional neural networks automatically extract multidimensional features images with high accuracy. To accurately identify different in complex backgrounds, an improved model (FCDM) based YOLOv5 is presented paper, which image-based data. By changing bounding box loss function SIoU Loss introducing directionality cost achieve faster convergence, training inference algorithm greatly improved. The Convolutional Block Attention Module (CBAM) introduced network fuse channel attention spatial improve recognition Path Aggregation Network (PANet) layer into a weighted Bi-directional Feature Pyramid (BiFPN) filter dimensions experimental results show that outperforms both performances. [email protected] detection, was by 3.9%, 4.0%, 3.8%, respectively. Among them, reached 83.1%, 90.6%. This indicates performance our proposed has been effectively some application prospects recognition.