基于YOLOv5改进的雾天行人与车辆检测算法

doi:10.16182/j.issn1004731x.joss.23-0725

摘要/Abstract

摘要：

由于在恶劣天气下汽车对环境的感知能力差，且对动态目标的检测能力有极大的影响，使得基于深度学习的目标检测网络在雾天行人车辆检测中出现精度低、鲁棒性差等问题。本文提出一种基于DehazeNet去雾算法与改进YOLOv5算法相结合的雾天检测方法—YOLOv5-SGE检测网络。通过取消初始锚框，实现锚框自适应计算，生成适合当前数据集的锚框；在特征提取模块加入三维权重注意力机制，使网络可以快速捕捉到感兴趣区域，抑制无用信息；使用轻量级卷积GSConv代替融合模块的标准卷积，弥补语义信息损失，减轻模型的复杂度；使用EIoU损失函数替换YOLOv5网络原损失函数CIoU，加快网络收敛速度。实验结果表明：所提算法具有较高的检测精度，mAP达到84%，验证了算法的有效性。

关键词: 自动驾驶, 图像去雾, YOLOv5, 轻量化卷积, 损失函数, 目标检测

Abstract:

Due to the poor environment perception of car in bad weather, the detection ability on dynamic targets is significantly reduced, and thus the problems such as low accuracy and poor robustness of the deep learning-based target detection network will occur when detecting pedestrians and vehicles in foggy days. A YOLOv5-SGE foggy detection network is proposed on the basis of the combination of image dehazing DehazeNet and the improved YOLOv5. The adaptive calculation of anchor frame is realized by canceling the initial anchor frame of YOLOv5, and the anchor frame suitable for the current dataset is generated. A three-dimensional weighted attention mechanism is added to the feature extraction module, so that the network can quickly capture the region of interest and suppress the useless information. Instead of the standard convolution of the fusion module the lightweight convolution GSConv is used to compensate for the loss of semantic information and reduce the complexity of tmodel. EIoU loss function is used to replace the original loss function CIoU of network to accelerate the network convergence speed. The experimental results show that, compared with the other four algorithms, the proposed model has a higher detection accuracy and mAP reaches 84%, which verifies the effectiveness of the proposed algorithm.

Key words: autonomous driving, image dehazing, YOLOv5, lightweight convolution, loss function, object detection

中图分类号:

TP391

苏彤,王颖,邓启扬等 . 基于YOLOv5改进的雾天行人与车辆检测算法[J]. 系统仿真学报, 2024, 36(10): 2413-2422.

Su Tong,Wang Ying,Deng Qiyang,et al . Improved Foggy Pedestrian and Vehicle Detection Algorithm Based on YOLOv5[J]. Journal of System Simulation, 2024, 36(10): 2413-2422.

图/表 10

图1

图2

表1

注意力机制结构设计和参数比较

模型	运算	参数量
SE	GAP，FC，ReLU	2 $C 2 / r$
CBAM	GAP，CAP，GMP，CMP，FC，ReLU，BN，C2D	2 $C 2 / r + 2 k 2$
ECA	GAP，C1D	K
SimAM	GAP，/，☉，+	0

表1

图3

图4

表2

表3

表4

图5

图6

参考文献 31

1	毛德乾, 高珊珊, 张晨昊, 等. 去光源影响的加权导向滤波图像去雾算法[J]. 计算机辅助设计与图形学学报, 2022, 34(2): 217-231.
	Mao Deqian, Gao Shanshan, Zhang Chenhao, et al. Image Dehazing via Weighted Guided Filtering with Removing Lighting Effects[J]. Journal of Computer-Aided Design & Computer Graphics, 2022, 34(2): 217-231.
2	王若萱, 吴建平, 徐辉. 自动驾驶汽车感知系统仿真的研究及应用综述[J]. 系统仿真学报, 2022, 34(12): 2507-2521.
	Wang Ruoxuan, Wu Jianping, Xu Hui. Overview of Research and Application on Autonomous Vehicle Oriented Perception System Simulation[J]. Journal of System Simulation, 2022, 34(12): 2507-2521.
3	张稀柳, 张晓玲, 何敏军. 基于改进YOLOX-s的车辆检测方法研究[J]. 系统仿真学报, 2024, 36(2): 487-496.
	Zhang Xiliu, Zhang Xiaoling, He Minjun. Research on Vehicle Detection Method Based on Improved YOLOX-s[J]. Journal of System Simulation, 2024, 36(2): 487-496.
4	董丽丽, 丁畅, 许文海. 基于直方图均衡化图像增强的两种改进方法[J]. 电子学报, 2018, 46(10): 2367-2375.
	Dong Lili, Ding Chang, Xu Wenhai. Two Improved Methods Based on Histogram Equalization for Image Enhancement[J]. Acta Electronica Sinica, 2018, 46(10): 2367-2375.
5	Suresh Chandra Raikwar, Tapaswi Shashikala, Chakraborty Soumendu. Bounding Function for Fast Computation of Transmission in Single Image Dehazing[J]. Multimedia Tools and Applications, 2022, 81(4): 5349-5372.
6	Ju Mingye, Ding Can, Ren Wenqi, et al. IDE: Image Dehazing and Exposure Using an Enhanced Atmospheric Scattering Model[J]. IEEE Transactions on Image Processing, 2021, 30: 2180-2192.
7	Narasimhan S G, Nayar S K. Vision and the Atmosphere[J]. International Journal of Computer Vision, 2002, 48(3): 233-254.
8	Li Shaoyi, Lin Jian, Yang Xi, et al. BPFD-Net: Enhanced Dehazing Model Based on Pix2pix Framework for Single Image[J]. Machine Vision and Applications, 2021, 32(6): 124.
9	He Kaiming, Sun Jian, Tang Xiaoou. Single Image Haze Removal Using Dark Channel Prior[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(12): 2341-2353.
10	Li Boyi, Peng Xiulian, Wang Zhangyang, et al. AOD-Net: All-in-one Dehazing Network[C]//2017 IEEE International Conference on Computer Vision (ICCV). Piscataway: IEEE, 2017: 4780-4788.
11	李永福, 崔恒奇, 朱浩, 等. 一种基于改进AOD-Net的航拍图像去雾算法[J]. 自动化学报, 2022, 48(6): 1543-1559.
	Li Yongfu, Cui Hengqi, Zhu Hao, et al. A Defogging Algorithm for Aerial Image with Improved AOD-Net[J]. Acta Automatica Sinica, 2022, 48(6): 1543-1559.
12	Cai Bolun, Xu Xiangmin, Jia Kui, et al. DehazeNet: An End-to-end System for Single Image Haze Removal[J]. IEEE Transactions on Image Processing, 2016, 25(11): 5187-5198.
13	郑凤仙, 王夏黎, 何丹丹, 等. 单幅图像去雾算法研究综述[J]. 计算机工程与应用, 2022, 58(3): 1-14.
	Zheng Fengxian, Wang Xiali, He Dandan, et al. Survey of Single Image Defogging Algorithm[J]. Computer Engineering and Applications, 2022, 58(3): 1-14.
14	Burges C J C. A Tutorial on Support Vector Machines for Pattern Recognition[J]. Data Mining and Knowledge Discovery, 1998, 2(2): 121-167.
15	Maria Dominic Savio M, Deepa T, Bonasu A, et al. Image Processing for Face Recognition Using HAAR, HOG, and SVM Algorithms[J]. Journal of Physics: Conference Series, 2021, 1964(6): 062023.
16	Thamilselvan P. Lung Cancer Prediction and Classification Using Adaboost Data Mining Algorithm[J]. International Journal of Computer Theory and Engineering, 2022, 14(4): 149-154.
17	Zheng Xuebin, Zhou Bingxin, Li Ming, et al. MathNet: Haar-like Wavelet Multiresolution Analysis for Graph Representation Learning[J]. Knowledge-Based Systems, 2023, 273: 110609.
18	徐艺博, 于清华, 王炎娟, 等. 基于多源信息融合的巡飞弹对地目标识别与毁伤评估[J]. 系统仿真学报, 2024, 36(2): 511-521.
	Xu Yibo, Yu Qinghua, Wang Yanjuan, et al. Ground Target Recognition and Damage Assessment of Patrol Missiles Based on Multi-source Information Fusion[J]. Journal of System Simulation, 2024, 36(2): 511-521.
19	Gao Chenqiang, Ye Sheng, Tian Hang, et al. Multi-scale Single-stage Pose Detection with Adaptive Sample Training in the Classroom Scene[J]. Knowledge-Based Systems, 2021, 222: 107008.
20	朱磊, 李东彪, 闫星志, 等. 基于改进Mask R-CNN深度学习算法的隧道裂缝智能检测方法[J]. 图学学报, 2023, 44(1): 177-183.
	Zhu Lei, Li Dongbiao, Yan Xingzhi, et al. Intelligent Detection Method of Tunnel Cracks Based on Improved Mask R-CNN Deep Learning Algorithm[J]. Journal of Graphics, 2023, 44(1): 177-183.
21	徐佳薇, 武杰, 雷宇, 等. 用于烟雾病检测的Faster RCNN改进算法[J]. 数据采集与处理, 2022, 37(6): 1391-1400.
	Xu Jiawei, Wu Jie, Lei Yu, et al. Improved Faster RCNN Algorithm for Moyamoya Disease Detection[J]. Journal of Data Acquisition and Processing, 2022, 37(6): 1391-1400.
22	Ren Shaoqing, He Kaiming, 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, 2017, 39(6): 1137-1149.
23	孙靖凯. 基于改进的AOD-net和YOLOv5的雾天车辆行人检测算法[D]. 哈尔滨: 哈尔滨理工大学, 2022.
	Sun Jingkai. Vehicleand Pedestrian Detection Algorithm Based on improved AOD-net and YOLOv5 in Hazy Eenvironment[D]. Harbin: Harbin University of Science and Technology, 2022.
24	Zhao Hang, Gallo O, Frosio I, et al. Loss Functions for Image Restoration with Neural Networks[J]. IEEE Transactions on Computational Imaging, 2017, 3(1): 47-57.
25	郭迎, 梁睿琳, 王润民. 基于CNN图像增强的雾天跨域自适应目标检测[J]. 计算机工程与应用, 2023, 59(16): 187-195.
26	陈琼红, 冀杰, 种一帆, 等. 基于AOD-net和SSD的雾天车辆和行人检测[J]. 重庆理工大学学报(自然科学), 2021, 35(5): 108-117.
	Chen Qionghong, Ji Jie, Chong Yifan, et al. Vehicle and Pedestrian Detection Based on AOD-net and SSD Algorithm in Hazy Environment[J]. Journal of Chongqing University of Technology(Natural Science), 2021, 35(5): 108-117.
27	汪昱东, 郭继昌, 王天保. 一种改进的雾天图像行人和车辆检测算法[J]. 西安电子科技大学学报, 2020, 47(4): 70-77.
	Wang Yudong, Guo Jichang, Wang Tianbao. Algorithm for Foggy-image Pedestrian and Vehicle Detection[J]. Journal of Xidian University, 2020, 47(4): 70-77.
28	Ding Changhao. Low Complexity and High Robustness Image Dehaze Algorithm[J]. Academic Journal of Computing & Information Science, 2022, 5(14): 27-32.
29	McCartney J E, Hall F F Jr. Optics of the Atmosphere: Scattering by Molecules and Particles[J]. Physics Today, 1977, 30(5): 76-77.
30	Yang Lingxiao, Zhang Ruyuan, Li Lida, et al. SimAM: A Simple, Parameter-free Attention Module for Convolutional Neural Networks[C]//Proceedings of the 38th International Conference on Machine Learning. Chia Laguna Resort: PMLR, 2021: 11863-11874.
31	Li H, Li J, Wei H, et al. Slim-neck by GSConv: A Better Design Paradigm of Detector Architectures for Autonomous Vehicles[J]. Journal of Real-time Image Processing, 2024, 21(3): 62-72.

去雾算法	薄雾	中雾	大雾
DCP	12.39	10.88	7.77
AOD-Net	12.19	13.68	14.66
DehazeNet	15.41	15.21	14.92

去雾算法	薄雾	中雾	大雾
DCP	0.75	0.68	0.63
AOD-Net	0.74	0.86	0.88
DehazeNet	0.85	0.90	0.88

算法	Person	Motorcycle	Car	Bicycle	mAP
AOD-YOLOv5	80.8	63.90	86.10	59.4	72.6
AOD-SSD	74.2	77.96	77.25	73.97	75.8
YOLOv8	86.2	69.60	89.00	79.20	81.0
YOLOv5-GS	85.6	76.50	86.10	74.80	80.8
YOLOv5-SGE	86.8	83.70	86.90	78.50	84.0

[1]	刘万军, 程裕茜, 曲海成. 基于生成对抗网络的图像自增强去雾算法[J]. 系统仿真学报, 2024, 36(5): 1093-1106.
[2]	史蓝兮, 颜文旭, 倪宏宇, 赵峰. 基于改进目标检测的动态场景SLAM研究[J]. 系统仿真学报, 2024, 36(4): 1028-1042.
[3]	韦金阳, 王科平, 杨艺, 费树岷. 基于多重迁移注意力的增量式图像去雾算法[J]. 系统仿真学报, 2024, 36(4): 969-980.
[4]	杨哲, 崔颖函, 郭灵犀, 李嘉鑫, 吴旭生. 融合数据仿真与深度学习算法的飞行器残骸搜寻技术[J]. 系统仿真学报, 2024, 36(10): 2238-2245.
[5]	董庆庆, 吴昊, 钱文华, 孔凤玲. 基于交叉细化和循环注意力的RGB-D显著性目标检测[J]. 系统仿真学报, 2023, 35(9): 1931-1947.
[6]	刘书刚, 张林坤, 杜昊东, 王洪涛. 雾天条件下改进YOLOv4的目标检测[J]. 系统仿真学报, 2023, 35(8): 1681-1691.
[7]	许仁杰, 张小明, 王晨, 吴鹏. 基于跨模块注意力的3D目标检测方法研究[J]. 系统仿真学报, 2023, 35(12): 2680-2691.
[8]	金炜东, 张述礼, 唐鹏, 张曼. 基于稠密残差块与通道像素注意力的图像去雾网络[J]. 系统仿真学报, 2022, 34(8): 1663-1673.
[9]	周培培, 侯幸林. 一种用于图像融合的无监督深度神经网络[J]. 系统仿真学报, 2022, 34(6): 1267-1274.
[10]	林士琪, 王纪凯, 裴浩渊, 赵皓, 陈宗海. 基于物体关系描述的单目语义SLAM方法[J]. 系统仿真学报, 2022, 34(2): 278-284.
[11]	朱志豪, 王艳, 纪志成. 基于模型压缩的安瓿瓶外观检测仿真研究[J]. 系统仿真学报, 2022, 34(12): 2575-2583.
[12]	王若萱, 吴建平, 徐辉. 自动驾驶汽车感知系统仿真的研究及应用综述[J]. 系统仿真学报, 2022, 34(12): 2507-2521.
[13]	尹智帅, 何嘉雄, 聂琳真, 管家意. 基于优化算法的自动驾驶车辆纵向自适应控制[J]. 系统仿真学报, 2021, 33(2): 409-420.
[14]	毕振波, 张世友, 杨花, 吴远红. 基于浅层机器学习的视频监控船舶检测综述[J]. 系统仿真学报, 2021, 33(12): 2792-2807.
[15]	石敏, 姚瀚钦, 李淳芃, 陈良臣. 基于深度Alignment网络的足部测量[J]. 系统仿真学报, 2020, 32(7): 1267-1278.