91 / 2025-04-14 10:48:26

GhostGD-YOLOv8: An efficient algorithm for forest fire detection by UAV images

YOLOv8, forest fire detection, unmanned aerial vehicle (UAV), lightweight characteristics, feature optimization

In this study, an improved YOLOv8 algorithm is designed specifically for forest fire detection task using unmanned aerial vehicle (UAV) images. In the backbone structure of YOLOv8, the GhostConv and C3Ghost modules are innovatively integrated to replace the original Conv and C2f modules. These modules possess excellent lightweight characteristics, which significantly reduce the computational load of the model while effectively improving the efficiency and quality of feature extraction. This enables the model to perform more robustly when processing complex forest scene images. Additionally, in the neck structure design, the Goldyolo architecture is employed for further optimization, which optimizes the feature transfer and fusion mechanism through its unique design, thereby enhancing the interaction between features across different scales. Experimental results demonstrate that the improved YOLOv8 algorithm exhibits superior performance in forest fire detection tasks. Compared with the original YOLOv8 model, improvements can be observed in detection accuracy, recall rate, and precision. The enhanced model can effectively address complex scenarios such as smoke occlusion and lighting variations in forest environments, providing robust technical support for early and accurate monitoring and warning of forest fires. It is anticipated to play a critical role in practical forest fire prevention efforts.