Accurately identifying the visibility levels of foggy images is of great significance for the early warning and prevention of fog disasters. Aiming at the problems of large training parameters and low recognition accuracy in current fog disaster warning models, a foggy image recognition model based on the improved VGG16 was proposed. This model took VGG16 as the basic architecture and incorporates the enhanced convolutional block attention module (ECBAM) and multi-loss functions to enhance model precision. Additionally, it replaced the three-layer fully connected layer with a two-layer structure and adds group normalization (GN) layers to accelerate model convergence. The optimized model reduced the number of parameters while improving training speed and recognition accuracy. Experimental results show these as follows. The improved model reduces the number of parameters by 52.22%, achieves a recognition accuracy of 88.96%, which is a 5.34% improvement over the original VGG16 model, and demonstrates higher recognition accuracy than comparative models such as AlexNet, VGG16, ResNet-50, and GoogLeNet. The recognition accuracies of the improved model for fog-free, light fog, dense fog, and thick fog images are 92.96%, 89.00%, 83.04%, and 81.76%, respectively, and the accuracy for extremely dense fog images is 73.75%. While the model meets daily operational needs, there is still room for optimization in recognizing extremely dense fog; Application tests using foggy videos from non-training periods show that the model can accurately identify fog disasters, with results consistent with those observed by visibility detection instruments. The proposed method can be applied to practical meteorological operations and holds high business application value.
ZHANG Guo-xing, MA Fang, LUO Jin, DENG Ji-bo, XU Ba-lin.
Fog Image Recognition and Application Based on Improved VGG16[J].
Science Technology and Engineering, 2026, 26(3): 934-942 DOI:10.12404/j.issn.1671-1815.2502431
基础模型采用计算机视觉领域的经典VGG16网络模型,将雾天图像作为模型的输入,经过一系列卷积池化处理,最后得到图像所属雾天类别的概率。VGG16模型结构如图2所示,该模型网络结构包含13个卷积层、5个池化层和3个全连接层。每个卷积层之后都使用ReLU(rectified linear unit)作为激活函数增加网络的非线性,池化层则是采用最大池化减少图像尺寸和保留图像特征。VGG16模型通过堆叠卷积层和池化层高效的提取图像特征,使其在图像分类任务中取得了较好的性能。
ZhangKe, HouJie. Research on image recognition method based on improved convolution neural network[J]. Science Technology and Engineering, 2020, 20(1): 252-257.
XiaoXiaomei, YangHongyun, YiWenlong, et al. Application of improved alexnet in image recognition of rice pests[J]. Science Technology and Engineering, 2021, 21(22): 9447-9454.
RenShujie, HuYong, HeWenxiang, et al. Microscopic image recognition of sandstone components based on deep learning[J]. Science Technology and Engineering, 2024, 24(9): 3727-3736.
LiYuechen, ZhaoXiao, WangRuonan, et al. Image recognition of Chinese herbal slices based on improved convolution neural network[J]. Science Technology and Engineering, 2024, 24(9): 3596-3604.
LiangXiaolong, LiJingang, XuPingping, et al. Leakage fault diagnosis of heating network based on CNN-Transformer[J]. Science Technology and Engineering, 2025, 25(13): 5589-5601.
ZhangZhenli, ChenYuan, FuHao, et al. A garbage imageclassification algorithm based on the improved EfficientNetV2 network[J]. Science Technology and Engineering, 2025, 25(10): 4229-4238.
MiaoKaichao, ZhouJianping, TaoPeng, et al. Visibility recognition of fog figure based on self-adaptive hybrid convolutional neural network[J]. Computer Engineering and Applications, 2020, 56(10): 205-212.
LiWeitao, HanHuihui, JiaoDian, et al. Research on intelligent cognition method of fog level based on deep transfer learning[J]. Journal of Electronic Measurement and Instrumentation, 2020, 34(2): 88-96.
ShenKecheng, ShiQuan, WangHan. Multimodal visibility deep learning model based on visible-infrared image pair[J]. Journal of Computer-Aided Design & Computer Graphics, 2021, 33(6): 939-946.
HuangLiang, ZhangZhendong, XiaoPengfei, et al. Classification and application of highway visibility based on deep learning[J]. Transactions of Atmospheric Sciences, 2022, 45(2): 203-211.
LiuDongwei, MuHaizhen, HeQianshan, et al. A low visibility recognition algorithm based on surveillance video[J]. Journal of Applied Meteorological Science, 2022, 33(4): 501-512.
TangWei, FangJianan, ZhangLong, et al. A method for measuring visibility under foggy weather for expressways based on siamese network[J]. Journal of Transport Information and Safety, 2023, 41(4): 122-131, 142.
YuLichun, LiuJinqing. Fog visibility measurement based on multi-task deep convolutional neural networks[J]. Journal of Southwest University Natural Science Edition, 2024, 46(4): 143-152.
[34]
ZhaoX, JiangJ, FengK, et al. The method of classifying fog level of outdoor video images based on convolutional neural networks[J]. Journal of the Indian Society of Remote Sensing, 2021, 49(9): 2261-2271.
[35]
LiL, YangB, ChenS. Detection of agglomerate fog based on a shallow convolutional neural network[J]. Multimedia Tools and Applications, 2022, 81(2): 2841-2857.
[36]
XiaJ, XuanD, TanL, et al. ResNet15: weather recognition on traffic road with deep convolutional neural network[J]. Advances in Meteorology, 2020(1): 6972826.
General Administration of Quality Supervision, Inspection and Quarantine of the People’s Republic of China, Standardization Administration of the People’s Republic of China.Forecast grades of fog: GB/T 27964—2011[S]. Beijing: China Standards Press, 2012.
[39]
WooS, ParkJ, LeeJ Y, et al. CBAM: convolutional block attention module[C]// Proceedings of the European Conference on Computer Vision (ECCV). Munich: Springer, 2018: 3-19.
[40]
WuY, HeK. Group normalization[C]// Proceedings of the European Conference on Computer Vision (ECCV). Munich: Springer, 2018: 3-19.
[41]
RossT Y, DollárG. Focal loss for dense object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2017: 2980-2988.
[42]
QiC, SuF.Contrastive-center loss for deep neural networks[C]//2017 IEEE International Conference on Image Processing (ICIP). New York: IEEE, 2017: 2851-2855.
[43]
HeK, ZhangX, RenS, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 770-778.
[44]
GoodfellowI, BengioY, CourvilleA. Deep learning[M]. Cambridge: MIT press, 2016.
[45]
HintonG E, SrivastavaN, KrizhevskyA, et al. Improving neural networks by preventing co-adaptation of feature detectors[J]. arXiv Preprint, arXiv: 1207. 0580, 2012.
[46]
SimonyanK. Very deep convolutional networks for large-scale image recognition[J]. arXiv prePrint, arXiv: 1409. 1556, 2014.
[47]
KrizhevskyA, SutskeverI, HintonG E. Imagenet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012, 25: 1097-1105.
[48]
HeK, ZhangX, RenS, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 770-778.
[49]
SzegedyC, LiuW, JiaY, et al. Going deeper with convolutions[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2015: 1-9.
[50]
DietterichT G. Ensemble methods in machine learning[C]// International Workshop on Multiple Classifier Systems. Berlin:Springer, 2000: 1-15.