基于CNN的林火检测和定位算法

打开文本图片集

中图分类号：TP183 文献标识码：A 文章编号：1674-0033(2025)06-0038-05

Forest Fire Detection and Localization Algorithm Based on CNN

ZHANG Xin

(College of Electronic Information and Electrical Engineering,Shangluo University，Shangluo 726000, Shaanxi)

Abstract:To improve the monitoring abilityof forest fires,images preprocessing isperformed based onthe characteristics of forest fire images,and the CNN-17 model is used to detect forest fire information within the images.The CNN-17 model adopts a "bottleneck structure"and uses an improved loss function as the loss function. When the algorithm detects a forest fire,it marks the fire and smoke area and locates the ignition point using geometric relationships. Comparing this method with SBP - YOLOv7model,FasterR （20 CNN model and DenseNet model. The results show that the CNN-17 model performs beter in key performance indicators such as accuracy,and can effectively detect flames and smoke simultaneously and locate the fire pointaccurately.Thismethodcan meet the needs of forest fire detection and location,and provide reference for forest fire early warning.

KeyWords:forest fire detection;CNN; loss function; ignition point localization

我国森林资源人均占有量偏低，单位面积蓄积量低于世界平均水平。（剩余9973字）