Low-light image enhancement method for underground mines based on an improved Zero-DCE model
-
摘要:
煤矿井下监控图像中存在噪声，清晰度低，且颜色和纹理信息缺失，采用基于机器学习的图像增强方法时还面临低照度−正常照度图像配对数据集采集困难的问题。提出一种改进零参考深度曲线估计（Zero−DCE）模型，并将其应用于矿井低照度图像增强。使用Leaky ReLU激活函数替换Zero−DCE模型中的ReLU激活函数，以加快模型收敛速度，提升低照度图像特征学习效率；在Zero−DCE模型浅层与深层网络之间的跳跃连接处引入卷积块注意力模块（CBAM），以提高模型对图像关键特征的表达能力；在浅层网络中引入非对称卷积块（ACB），以优化模型对局部图像特征的学习能力和细节特征的表现能力；在深层网络中采用串联卷积核（CCK），以降低模型参数量和计算量，缩短模型训练时间。采用LOL公共数据集和矿井自建数据集进行实验验证，结果表明：改进Zero−DCE模型的均方误差（MSE）、峰值信噪比（PSNR）、结构相似性（SSIM）、自然图像质量评估器（NIQE）和视觉信息保真度（VIF）整体上优于典型图像增强模型，在自建数据集上的MSE和NIQE较Zero−DCE模型分别降低16.25%和2.93%，PSNR，SSIM和VIF分别提高2.87%，1.87%和17.64%；图像增强视觉效果较好，可在提高图像亮度的同时有效保留细节纹理信息，降噪效果明显；对单幅图像的推理时间为0.138 s，可实现图像实时增强。
-
关键词:
- 矿井低照度图像 /
- 图像增强 /
- 零参考深度曲线估计网络 /
- Zero−DCE模型 /
- 无监督学习
Abstract:Underground coal mine surveillance images suffer from noise, low clarity, missing color, and texture information. Additionally, machine learning-based image enhancement methods face challenges in collecting paired low-light and normal-light image datasets. To address these issues, this paper proposes an improved Zero-Reference Deep Curve Estimation (Zero-DCE) model for enhancing low-light images in mines. The ReLU activation function in the Zero-DCE model was replaced with Leaky ReLU to accelerate model convergence and improve the efficiency of low-light image feature learning. A Convolutional Block Attention Module (CBAM) was introduced at the skip connections between the shallow and deep networks of the Zero-DCE model to enhance the model's ability to capture key image features. An Asymmetric Convolution Block (ACB) was incorporated into the shallow network to optimize the model's learning of local image features and its ability to represent fine details. A Cascaded Convolution Kernel (CCK) was employed in the deep network to reduce the number of model parameters and computational cost, thereby shortening the training time. Experimental validation was conducted using the LOL public dataset and a self-built mine dataset. The results showed that the improved Zero-DCE model outperformed typical image enhancement models in terms of Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), Structural Similarity (SSIM), Natural Image Quality Evaluator (NIQE), and Visual Information Fidelity (VIF). Specifically, on the self-built dataset, MSE and NIQE decreased by 16.25% and 2.93%, respectively, while PSNR, SSIM, and VIF increased by 2.87%, 1.87%, and 17.64%, respectively. The enhanced images exhibited superior visual quality, effectively improving brightness while preserving detailed texture information and significantly reducing noise. The inference time for a single image was 0.138 seconds, enabling real-time image enhancement.
-
-
![]()
图 4 基于改进Zero−DCE模型的矿井低照度图像增强方法流程
Figure 4. Process of low-light coal mine image enhancement method based on improved Zero-DCE model
表 1 消融实验指标对比
Table 1 Indicators comparison of ablation experiment
模型 MSE PSNR SSIM NIQE VIF 模型1 310.51 23.21 0.90 8.52 0.32 模型2 302.75 23.32 0.91 8.49 0.32 模型3 300.66 23.35 0.90 8.50 0.36 模型4 275.47 23.73 0.91 8.51 0.35 模型5 311.95 23.19 0.90 8.52 0.32 模型6 291.13 23.49 0.90 8.33 0.38 模型7 262.47 23.94 0.92 8.29 0.38 本文模型 260.06 23.98 0.92 8.27 0.38 
下载: 导出CSV
表 2 LOL数据集图像增强指标对比
Table 2 Comparison of image enhancement indicators for LOL dataset
模型 MSE PSNR SSIM NIQE VIF LIME 2 618.67 13.95 0.69 9.41 0.26 Retinex−Net 1 367.98 16.77 0.79 9.76 0.22 KinD 1 117.07 17.65 0.88 7.56 0.29 KinD Plus 739.74 19.44 0.89 9.41 0.29 Zero−DCE 1 069.25 17.84 0.88 9.90 0.27 EnlightenGAN 1 161.66 17.48 0.86 9.70 0.30 SSIENet 731.27 18.94 0.91 10.53 0.33 本文模型 776.39 19.23 0.91 9.28 0.37
下载: 导出CSV
表 3 矿井数据集图像增强指标对比
Table 3 Comparison of image enhancement indicators for mine dataset
模型 MSE PSNR SSIM NIQE VIF LIME 992.84 18.16 0.72 9.30 0.31 Retinex−Net 1742.93 15.72 0.74 9.27 0.34 KinD 318.48 23.10 0.94 9.29 0.37 KinD Plus 461.40 21.49 0.89 10.58 0.32 Zero−DCE 310.51 23.21 0.90 8.52 0.32 EnlightenGAN 322.167 23.05 0.90 8.73 0.31 SSIENet 314.109 23.16 0.91 8.62 0.32 本文模型 260.06 23.98 0.92 8.27 0.38
下载: 导出CSV
表 4 各模型性能对比
Table 4 Performance comparison of various models
模型 推理时间/s 参数量/106个 计算量/109个 LIME 15.467 — — Retinex−Net 0.704 0.555 587.470 KinD 1.033 8.160 574.954 KinD Plus 12.716 8.275 12 238.026 Zero−DCE 0.123 0.079 84.990 EnlightenGAN 0.313 8.637 273.240 SSIENet 0.797 0.682 154.665 本文模型 0.138 0.052 55.738
下载: 导出CSV
-
[1] 程德强,钱建生,郭星歌,等. 煤矿安全生产视频AI识别关键技术研究综述[J]. 煤炭科学技术,2023,51(2):349-365. CHENG Deqiang,QIAN Jiansheng,GUO Xingge,et al. Review on key technologies of AI recognition for videos in coal mine[J]. Coal Science and Technology,2023,51(2):349-365.
[2] 范伟强,刘毅. 基于自适应小波变换的煤矿降质图像模糊增强算法[J]. 煤炭学报,2020,45(12):4248-4260. FAN Weiqiang,LIU Yi. Fuzzy enhancement algorithm of coal mine degradation image based on adaptive wavelet transform[J]. Journal of China Coal Society,2020,45(12):4248-4260.
[3] 吴佳奇,张文琪,陈伟,等. 基于改进CycleGAN的煤矿井下低照度图像增强方法[J]. 华中科技大学学报(自然科学版),2023,51(5):40-46. WU Jiaqi,ZHANG Wenqi,CHEN Wei,et al. Image enhancement method of underground low illumination in coal mine based on improved CycleGAN[J]. Journal of Huazhong University of Science and Technology(Natural Science Edition),2023,51(5):40-46.
[4] 程健,李昊,马昆,等. 矿井视觉计算体系架构与关键技术[J]. 煤炭科学技术,2023,51(9):202-218. CHENG Jian,LI Hao,MA Kun,et al. Architecture and key technologies of coalmine underground vision computing[J]. Coal Science and Technology,2023,51(9):202-218.
[5] 田会娟,蔡敏鹏,关涛,等. 基于YCbCr颜色空间的Retinex低照度图像增强方法研究[J]. 光子学报,2020,49(2):173-184. TIAN Huijuan,CAI Minpeng,GUAN Tao,et al. Low-light image enhancement method using Retinex method based on YCbCr color space[J]. Acta Photonica Sinica,2020,49(2):173-184.
[6] GUO Xiaojie,LI Yu,LING Haibin. LIME:low-light image enhancement via illumination map estimation[J]. IEEE Transactions on Image Processing :a Publication of the IEEE Signal Processing Society,2016,26(2):982-993.
[7] WEI Chen,WANG Wenjing,YANG Wenhan,et al. Deep Retinex decomposition for low-light enhancement[Z/OL]. [2024-10-11]. https://doi.org/10.48550/arXiv.1808.04560.
[8] ZHANG Yonghua,ZHANG Jiawan,GUO Xiaojie. Kindling the darkness:a practical low-light image enhancer[C]. The 27th ACM International Conference,Torino, 2019. DOI: 10.1145/3343031.3350926.
[9] ZHANG Yonghua,GUO Xiaojie,MA Jiayi,et al. Beyond brightening low-light images[J]. International Journal of Computer Vision,2021,129(4):1-25.
[10] 王满利,张航,李佳悦,等. 基于深度神经网络的煤矿井下低光照图像增强算法[J]. 煤炭科学技术,2023,51(9):231-241. WANG Manli,ZHANG Hang,LI Jiayue,et al. Deep neural network-based image enhancement algorithm for low-illumination images underground coal mines[J]. Coal Science and Technology,2023,51(9):231-241.
[11] GUO Chunle,LI Chongyi,GUO Jichang,et al. Zero-reference deep curve estimation for low-light image enhancement[C]. The IEEE/CVF Conference on Computer Vision and Pattern Recognition,Seattle,2020:1777-1786.
[12] JIANG Yifan,GONG Xinyu,LIU Ding,et al. EnlightenGAN:deep light enhancement without paired supervision[J]. IEEE Transactions on Image Processing:a Publication of the IEEE Signal Processing Society,2021,30:2340-2349. DOI: 10.1109/TIP.2021.3051462
[13] ZHANG Yu,DI Xiaoguang,ZHANG Bin,et al. Self-supervised image enhancement network:training with low light images only[Z/OL]. [2024-10-11]. https://doi.org/10.48550/arXiv.2002.11300.
[14] 田子建,阳康,吴佳奇,等. 基于LMIENet图像增强的矿井下低光环境目标检测方法[J]. 煤炭科学技术,2024,52(5):222-235. TIAN Zijian,YANG Kang,WU Jiaqi,et al. LMIENet enhanced object detection method for low light environment in underground mines[J]. Coal Science and Technology,2024,52(5):222-235.
[15] ZHANG Jian,CRAIGMILE P F,CRESSIE N. Loss function approaches to predict a spatial quantile and its exceedance region[J]. Technometrics,2008,50(2):216-227. DOI: 10.1198/004017008000000226
[16] YASHWANT K,VIJAYSHRI C. An image fusion approach based on adaptive fuzzy logic model with local level processing[J]. International Journal of Computer Applications,2015,124(1):39-42. DOI: 10.5120/ijca2015905317
[17] BERLOFFA G,MODENA F. Income shocks,coping strategies,and consumption smoothing:an application to Indonesian data[J]. Journal of Asian Economics,2013,24:158-171. DOI: 10.1016/j.asieco.2012.11.004
[18] LI Zhengguo,ZHENG Jinghong,RAHARDJA S. Detail-enhanced exposure fusion[J]. IEEE Transactions on Image Processing,2012,21(11):4672-4676. DOI: 10.1109/TIP.2012.2207396
[19] XU Bing,WANG Naiyan,CHEN Tianqi,et al. Empirical evaluation of rectified activations in convolutional network[Z/OL]. [2024-10-11]. https://doi.org/10.48550/arXiv.1505.00853.
[20] DING Xiaohan, GUO Yuchen, DING Guiguang, et al. ACNet: strengthening the kernel skeletons for powerful CNN via asymmetric convolution blocks[J]. International Conference on Computer Vision, Seoul, 2019. DOI: 10.1109/ICCV.2019.00200.
[21] WOO S,PARK J,LEE J Y,et al. CBAM:convolutional block attention module[C]. European Conference on Computer Vision,Munich,2018:3-19.
[22] 苗作华,赵成诚,朱良建,等. 矿井井下非均匀照度图像增强算法[J]. 工矿自动化,2023,49(11):92-99. MIAO Zuohua,ZHAO Chengcheng,ZHU Liangjian,et al. Image enhancement algorithm for non-uniform illumination in underground mines[J]. Journal of Mine Automation,2023,49(11):92-99.
[23] 赵征鹏,李俊钢,普园媛. 基于卷积神经网络的Retinex低照度图像增强[J]. 计算机科学,2022,49(6):199-209. ZHAO Zhengpeng,LI Jungang,PU Yuanyuan. Low-light image enhancement based on Retinex theory by convolutional neural network[J]. Computer Science,2022,49(6):199-209.
[24] SETIADI D R I M. PSNR vs SSIM:imperceptibility quality assessment for image steganography[J]. Multimedia Tools and Applications,2020,80(6):1-22.
-
期刊类型引用(19)
1. 郭爱军. 基于ELM-VPMCD的输送带超声缺陷识别分析. 工矿自动化. 2024(S1): 11-14+33 . 
本站查看
2. 马孝威，李标. 基于云边协同的带式输送机故障诊断技术. 工矿自动化. 2024(S1): 119-121+150 . 本站查看
3. 王磊. 基于带式输送机的驱动技术优化研究. 机械管理开发. 2023(01): 176-177+180 . 百度学术
4. 史春鹏. 带式输送机驱动滚筒轴承故障诊断及试验. 机械工程与自动化. 2023(01): 152-153+156 . 百度学术
5. 李标. 带式输送机多电动机驱动系统转速同步控制方法. 工矿自动化. 2023(05): 112-119 . 本站查看
6. 易辉. 矿用带式输送机噪声源分析及降噪措施. 陕西煤炭. 2023(04): 91-94 . 百度学术
7. 马孝威. 基于多源信息融合的托辊故障检测系统. 煤矿机械. 2023(09): 168-170 . 百度学术
8. 翟建忠. 矿用带式输送机卸载滚筒机架振动问题分析与改进研究. 机械管理开发. 2023(08): 157-158+165 . 百度学术
9. 许良雨. 带式输送机滚筒轴承故障预警装置研究. 煤. 2022(07): 54-55 . 百度学术
10. 吴文臻，程继明，李标. 矿用带式输送机托辊音频故障诊断方法. 工矿自动化. 2022(09): 25-32 . 本站查看
11. 刘媛媛. 煤矿机电设备智能化维护研究现状与发展趋势. 工矿自动化. 2021(07): 79-84 . 本站查看
12. 杨战社，孔晨再，荣相，魏礼鹏，史小军. 基于EEMD能量熵与ANN的矿用异步电机故障诊断. 微电机. 2021(08): 23-27+61 . 百度学术
13. 韩越. 带式输送机驱动滚筒轴承故障特征提取分析研究. 煤矿机械. 2021(10): 162-165 . 百度学术
14. 张猛，苗长云，孟德军. 轴承早期故障特征提取方法研究. 工矿自动化. 2020(04): 85-90+116 . 本站查看
15. 赵新哲. 管状带式输送机的调整和维护. 海峡科技与产业. 2020(03): 67-70 . 百度学术
16. 杨东明. 综采工作面转载机与带式输送机联锁控制改造. 机电工程技术. 2020(06): 211-213 . 百度学术
17. 王文杰. 矿用带式输送机的非线性动力学建模研究. 机械制造与自动化. 2020(05): 131-134 . 百度学术
18. 杨志彦. 掘进巷道带式输送机防跑偏振动装置设计应用. 机械管理开发. 2019(07): 39-40 . 百度学术
19. 陈鑫. 带式输送机改向滚筒的改进和有限元分析. 机械管理开发. 2019(09): 140-142 . 百度学术
其他类型引用(3)
计量
- 文章访问数: 79
- HTML全文浏览量: 15
- PDF下载量: 17
- 被引次数: 22
