摘要
为了解决传统火焰烟雾检测算法在光照条件恶劣和山林云雾影响的条件下存在漏检误检、准确性下降的缺陷,提出了一种基于YOLOv5s目标检测框架的森林烟火检测算法。首先,针对火焰烟雾目标特征复杂的问题,在C3模块中融合Res2Net,增强了网络在不同尺度下的特征表示能力。然后,在主干检测网络加入SE注意力模块,达到抑制干扰信息,提升模型表征能力的效果。最后,通过集成GIOU优化损失函数,进一步提高了检测精度。改进后的的算法相比于传统算法,mAP50值提高了1.8%,P值提高了0.9%,R值提高了0.6%。
In order to solve the defects of traditional flame smoke detection algorithms in terms of leakage and misdetection and accuracy degradation under the conditions of poor lighting conditions and the influence of clouds and fog in mountain forests, a forest smoke and fire detection algorithm based on the YOLOv5 target detection framework is proposed. Firstly, to address the problem of complex features of flame smoke targets, Res2Net is fused in the C3 module, which enhances the feature representation ability of the network at different scales. Then, the SE attention module is added to the backbone detection network to achieve the effect of suppressing the interference information and enhancing the model representation ability. Finally, the detection accuracy is further improved by integrating GIOU to optimize the loss function. The improved algorithm continues to improve the mAP50 value by 1.8%, the P value by 0.9%, and the R value by 0.6% compared with the traditional algorithm.
出处
《计算机科学与应用》
2024年第4期290-297,共8页
Computer Science and Application
