| 734 | 9 | 214 |
| 下载次数 | 被引频次 | 阅读次数 |
由于复杂环境下类烟火物体的干扰，常导致火灾检测误判。为了提高图像中火灾信号的检测精度，减少火灾误报，利用传统光谱分析在火灾图像检测技术中的优势，提出了一种基于小波变换的YOLOv5火灾检测改进算法。该算法利用二维Haar小波变换提取图像的光谱特征，将其输入到YOLOv5s的主干网络CSPDarknet中，与卷积层进行通道上的特征融合，增强烟火的纹理细节特征；通过嵌入CA注意力机制的CAC3模块，对融合小波特征后的网络层的位置信息进行增强，提高网络的信息提取和定位能力；为明确衡量边界框宽高的真实差，平衡烟火难易样本，采用α-EIOU损失函数替换原本的CIOU,提高框定位准确性。在公开的火灾数据基础上结合自制火灾数据构建火灾数据集，并进行模型训练和推理。实验结果表明，改进后算法的mAP比原YOLOv5s提升了2.3%,实现了对火灾场景烟火目标较好的检测效果。
Abstract：Due to the interference of pyrotechnic objects in complex environment, misjudgment of fire detection is often caused. In order to improve the detection accuracy of fire signals in images and reduce false alarms, an improved YOLOv5 fire detection algorithm based on wavelet transform is proposed by taking advantage of traditional spectral analysis in fire detection technology. The algorithm uses two-dimensional Haar wavelet transform to extract the spectral features of the image and input them into CSPDarknet, the backbone network of YOLOv5s, and carry out feature fusion on the channel with the convolution layer to enhance the texture detail features of the fireworks. The location information of the network layer is enhanced by the CAC3 module embedded with CA attention mechanism, and the information extraction and positioning ability of the network is improved. In order to measure the difference of the width and height of the bounding box and balance difficulty samples of fireworks, the original CIOU is replaced by the α-EIOU loss function, which improves the frame orientation accuracy. Based on the open fire data and homemade fire data, the fire data set is constructed, and the model training and reasoning are carried out. The experimental results show that the mAP of the improved algorithm is 2.3% higher than that of the original YOLOv5s, achieving a better detection effect on the firework target in the fire scene.
[1] 祝玉华，司艺艺，李智慧.基于深度学习的烟雾与火灾检测算法综述[J].计算机工程与应用，2022,58(23):1-11.
[2] FRIZZI S,KAABI R,BOUCHOUICHAM,et al.Convolutional Neural Network for Video Fire and Smoke Detection[C]//IECON 2016-42nd Annual Conference of the IEEE Industrial Electronics Society.Florence:IEEE,2016:877-882.
[3] REN S Q,HE K M,GIRSHICK R,et al.Faster R-CNN:Towards Real-time Object Detection with Region Proposal Networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,39(6):1137-1149.
[4] REDMON J,FARHADI A.YOLOv3:An Incremental Improvement[J/OL].(2018-04-08)[2023-03-11].https://arxiv.org/abs/1804.02767.
[5] ZHANG S F,WEN L Y,BIAN X,et al.Single-shot Refinement Neural Network for Object Detection[C]//2018 IEEE Conference on Computer Vision and Pattern Recognition.Salt Lake City:IEEE,2018:4203-4212.
[6] LIU W,ANGUELOV D,ERHAN D,et al.SSD:Single Shot Multibox Detector[J/OL].(2015-12-08)[2023-03-11].https://arxiv.org/abs/1512.02325.
[7] LIN G H,ZHANG Y M,XU G,et al.Smoke Detection on Video Sequences Using 3D Convolutional Neural Networks[J].Fire Technology,2019,55(5):1827-1847.
[8] QIAN H M,SHI F,CHEN W,et al.A Fire Monitoring and Alarm System Based on Channel-wise Pruned YOLOv3[J].Multimedia Tools and Applications,2021,81(2):1833-1851.
[9] 王恒涛，张上，张朝阳，等.基于YOLOv5的轻量化PCB缺陷检测[J].无线电工程，2022,52(11):2094-2100.
[10] 皮骏，刘宇恒，李久昊.基于YOLOv5s的轻量化森林火灾检测算法研究[J].图学学报，2023,44(1):26-32.
[11] CHEN Z C,YANG J,CHEN L F,et al.Garbage Classification System Based on Improved ShuffleNet v2[J].Resources,Conservation and Recycling,2022,78:106090.
[12] HU Y W,ZHAN J L,ZHOU G X,et al.Fast Forest Fire Smoke Detection Using MVMNet[J].Knowledge-Based Systems,2022,241:108219.
[13] XU R J,LIN H F,LU K J,et al.A Forest Fire Detection System Based on Ensemble Learning[J].Forests,2021,12(2):217.
[14] ZHANG X B,QIAN K,JING K H,et al.Fire Detection Based on Convolutional Neural Networks with Channel Attention[C]//2020 Chinese Automation Congress (CAC).Shanghai:IEEE,2020:3080-3085.
[15] HOU Q B,ZHOU D Q,FENG J S.Coordinate Attention for Efficient Mobile Network Design[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Nashville:IEEE,2021:13713-13722.
[16] HU J,SHEN L,SUN G.Squeeze-and-Excitation Networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2020,42(8):2011-2023.
[17] WOO S,PARK J,LEE J Y,et al.CBAM:Convolutional Block Attention Module[J/OL].(2018-07-18)[2023-03-11].https://arxiv.org/pdf/1807.06521.pdf.
[18] ZHANG Y F,REN W Q,ZHANG Z,et al.Focal and Efficient IOU Loss for Accurate Bounding Box Regression[J].Neurocomputing,2022,506:146-157.
[19] HE J B,ERFANI S,MA X J,et al.AlphaIoU:A Family of Power Intersection over Union Losses for Bounding Box Regression[J/OL].(2022-01-02)[2023-03-11].https://arxiv.org/abs/2110.13675.
基本信息:
中图分类号:X932;TP183;TP391.41
引用信息:
[1]章曙光,唐锐,邵政瑞,等.一种基于小波变换的YOLOv5火灾检测改进算法[J].无线电工程,2023,53(10):2303-2310.
基金信息:
安徽省高校协同创新项目(GXXT-2021-093); 安徽省重点研究与开发计划项目(202004a07020050); 基于国产智能芯片的宿舍安全管理平台应用开发服务(HYB20200190)~~