Research on Forest Fire Image Recognition Based on Improved YOLOv5

Forest fire image recognition based on YOLOv5 has great development and application prospects in the field of forest fire prevention today, among which the target recognition accuracy of early forest fire images is the key. In response to problem of low accuracy in image recognition during the early stages of forest fires, this article proposes an improved YOLOv5 algorithm based on YOLOv5 algorithm, which is improved on YOLOv5 backbone network to achieve better accuracy in image recognition during the early stages of forest fires. Insert CBAM into C3 module in YOLOv5 backbone network and add BotNET module at the end of backbone network. A comparative experiment was conducted between the improved YOLOv5 and the original YOLOv5, experimental results showed that improved YOLOv5 improved the average detection accuracy in both small fire target images and fire target image features that were not obvious. This can prove that improved YOLOv5 method proposed in this paper has a certain effect on early image recognition of forest fires.