一种改进YOLOv5s的森林火灾烟雾检测算法OA北大核心CSTPCD
An Improved YOLOv5s Forest Fire Smoke Detection Algorithm
提出一种基于改进YOLOv5s的森林火灾烟雾检测算法.构建包含16573幅图片的火焰烟雾数据集,解决训练数据不足的问题,提高训练模型的泛化能力.设计一种轻量化的GC-C3模块替换原有的C3模块,减少模型参数量和计算量;将加权双向特征金字塔网络结构融合到Neck结构中,增强网络对于中小目标的检测能力;修改网络空间金字塔池化结构,使用SimSPPF结构替换SPPF,提高了网络的计算效率和检测准确度;将边界框回归损失函数CIOU替换为Focal-EIOU,加快模型的收敛速度,解决正负样本不匹配的问题.实验结果表明:改进之后的网络平均检测准确度提高2.3%,模型参数数量下降46.7%,模型计算量下降47.5%.
A forest fire smoke detection algorithm based on improved YOLOv5s is proposed.A fire and smoke dataset containing 16573 images is constructed to solve the problem of insufficient training data sets and improve the generalization ability of the training model.A lightweight GC-C3 module is designed to replace the original C3 module and reduce the number of model parameters and calculation.The weighted bidirectional feature pyramid network is integrated into the Neck structure to enhance the detection ability of the network for small and medium targets.The network space pyramid pool structure is modified,SPPF is replaced by SimSPPF structure,and the computing efficiency and detection accuracy of the network are improved.The bounding box regression loss function CIOU is replaced by Focal-EIOU to accelerate the convergence of the model and solve the problem of mismatch between positive and negative samples.The experimental results show that the average detection accuracy of the improved network is increased by 2.3%,the number of model parameters is decreased by 46.7%,and the calculation amount of the model is decreased by 47.5%.
张立国;张琦;金梅;袁煜淋;王泓沣
燕山大学电气工程学院,河北秦皇岛 066004
机器视觉火灾烟雾检测深度学习YOLOv5s轻量化小目标检测Focal-EIOU
machine visionfire smoke detectiondeep learningYOLOv5slight weightsmall target detectionFocal-EIOU
《计量学报》 2024 (009)
1314-1323 / 10
河北省中央引导地方专项(199477141G)
评论