| 摘 要: 随着全球天气持续变暖,高温、干旱、大风等极端天气呈现多发并发态势,导致全球范围内森林火灾频繁爆发。为了提升森林火灾检测精确率和实时性,提出了一种基于改进YOLOv8的森林火灾检测算法模型。该模型在Neck端上采样阶段中的Upsample结构后面以及每个C2F 模块后面添加CBAM(Convolutional Block Attention Module)注意力机制,通过对特征通道和空间的学习,提高模型的特征提取能力,解决火焰和烟雾图像分辨率低和受不同场景因素影响导致的识别率低的问题。将该算法模型应用于火灾数据集进行训练、验证、测试发现,与原算法模型相比,经改进的算法模型的准确率和召回率分别提高了6.5%和6.8%,其中mAP@0.5提高了4.8%。实验结果表明,改进后的算法模型能够实现对森林火灾的实时监测与精确识别。 |
| 关键词: YOLOv8;CBAM注意力机制;森林火灾检测 |
|
中图分类号: TP391
文献标识码: A
|
|
| Research on Forest Fire Detection Algorithm Based on Improved YOLOv8 |
|
CHEN Yixiao, SHEN Jingfeng, ZHONG Liangwei
|
(School of Mechanical Engineering, University of Shanghai f or Science and Technology, Shanghai 200093, China)
cyx1873891950@163.com;; sh_jf@163.com; zlv@usst.edu.cn
|
| Abstract: With the continuous warming of global weather, extreme weather such as high temperatures, droughts, and strong winds are becoming more frequent and concurrent, leading to frequent outbreaks of forest fires around the world. In order to improve the accuracy and real-time performance of forest fire detection, this paper proposes a forest fire detection algorithm model based on improved YOLOv8. This model adds CBAM (Convolutional Block Attention Module) attention mechanism after the Upsample structure in the Neck upsampling stage and after each C2F module, which improves the feature extraction ability of the model through learning feature channels and spaces and solves the problem of low resolution of flame and smoke images and low recognition rate caused by different scene factors. Applying the algorithm model to the fire dataset for training, validation, and testing, it has been found that compared with the original algorithm model, the accuracy and recall rates of the improved algorithm model increase by 6.5% and 6.8% , respectively and mAP @ 0.5 increases by 4.8% . The experimental results show that the improved algorithm model can achieve real-time monitoring and accurate identification of forest fires. |
| Keywords: YOLOv8; CBAM attention mechanism; forest fire detection |
