面向煤矿井下低光照图像的增强方法

孔二伟; 张亚邦; 李佳悦; 王满利

doi:10.13272/j.issn.1671-251x.2022110054

面向煤矿井下低光照图像的增强方法

doi: 10.13272/j.issn.1671-251x.2022110054

1.
平顶山天安煤业股份有限公司, 河南平顶山　467000
2.
河南理工大学物理与电子信息学院, 河南焦作　454000

基金项目: 国家自然科学基金项目（52074305）；河南理工大学博士基金项目（B2021-64)。

详细信息

作者简介:
孔二伟（1982—），男，河南平顶山人，高级工程师，硕士，现主要从事电气工程方面的工作，E-mail：kongerwei1982@163.com

通讯作者:
李佳悦（1999—），女，山西运城人，硕士研究生，研究方向为图像处理、深度学习，E-mail：lijiayue0827@163.com

中图分类号: TD67
计量
- 文章访问数: 465
- HTML全文浏览量: 205
- PDF下载量: 87
- 被引次数: 0
出版历程
- 收稿日期: 2022-11-14
- 修回日期: 2023-04-20
- 网络出版日期: 2023-04-27

An enhancement method for low light images in coal mines

1.
Pingdingshan Tian'an Coal Mining Co., Ltd., Pingdingshan 467000, China
2.
School of Physics and Electronic Information Engineering, Henan Polytechnic University, Jiaozuo 454000, China

摘要

摘要: 煤矿井下照明有限，并且具有大量粉尘、雾气，使得采集到的图像对比度低、光照不均、细节信息弱，并含有大量噪声。基于传统模型的图像增强方法鲁棒性较差，常会引起图像过度增强和色彩失真；基于深度学习的图像增强方法大多没有考虑增强引起的噪声放大。针对上述问题，提出了一种面向煤矿井下低光照图像的增强方法。采用卷积神经网络构建图像增强网络，该网络包括特征提取模块、增强模块和融合模块。特征提取模块对输入图像进行不同程度的卷积，提取多层次的图像特征，得到多个特征层；增强模块对提取到的特征层通过子网络进行增强，强化不同程度的细节特征；融合模块将增强后的特征层进行融合，输出增强图像。之后通过结构损失函数、内容损失函数和区域损失函数的约束，提高图像质量并有效抑制图像颜色失真与噪声放大，得到最终的增强图像。实验结果表明，该方法能够有效提升煤矿井下低光照图像的亮度和对比度，并且具有较强的噪声抑制能力，使图像能更好地恢复原有的细节信息，同时避免出现过曝光或颜色失真。
- 煤矿低光照图像 /
- 图像增强 /
- 图像去噪 /
- 深度学习 /
- 损失函数
Abstract: Underground lighting in coal mines is limited. There is a large amount of dust and mist, resulting in low contrast, uneven lighting, weak detail information, and a large amount of noise in the collected images. The image enhancement methods based on traditional models have poor robustness, often causing excessive image enhancement and color distortion. Most image enhancement methods based on deep learning do not consider the noise amplification caused by enhancement. In order to solve the above problems, an enhancement method for low light images in coal mines is proposed. The image enhancement network is constructed by using convolutional neural networks. The network includes feature extraction modules, enhancement modules, and fusion modules. The feature extraction module convolves the input image to varying degrees, extracts multi-level image features, and obtains multiple feature layers. The enhancement module enhances the extracted feature layers through sub-networks to enhance different levels of detail features. The fusion module fuses the enhanced feature layers and outputs enhanced images. Then, through the constraints of the structure loss function, content loss function and area loss function, the image quality is improved. The image color distortion and noise amplification are effectively suppressed to obtain the final enhanced image. The experimental results show that this method can effectively improve the brightness and contrast of low light images in coal mines. The method has strong noise suppression capability, enabling the image to better restore the original details while avoiding overexposure or color distortion.
- low light images of coal mines /
- image enhancement /
- image denoising /
- deep learning /
- loss function

HTML全文

图 1 面向煤矿井下低光照图像的增强网络整体结构

Figure 1. Overall structure of enhancement network for underground coal mine low-light image

下载: 全尺寸图片幻灯片

图 2 FEM结构

Figure 2. Structure of feature extraction module

下载: 全尺寸图片幻灯片

图 3 EM结构

Figure 3. Structure of enhancement module

下载: 全尺寸图片幻灯片

图 4 EM子网络结构

Figure 4. Sub-network structure of enhancement module

下载: 全尺寸图片幻灯片

图 5 FM结构

Figure 5. Structure of fusion module

下载: 全尺寸图片幻灯片

图 6 本文方法增强效果

Figure 6. Enhancement effect of the proposed method

下载: 全尺寸图片幻灯片

图 7 不同方法下图像增强效果对比

Figure 7. Comparison of image enhancement effect under different methods

下载: 全尺寸图片幻灯片

图 8 不同方法下图像客观评价结果对比

Figure 8. Comparison of image objective evaluation results under different methods

下载: 全尺寸图片幻灯片

图 9 不同特征层数下图像增强效果对比

Figure 9. Comparison of image enhancement effect under different feature layers

下载: 全尺寸图片幻灯片

图 10 EM添加Concat层前后增强效果对比

Figure 10. Comparison of enhancement effect before and after enhancement module adding Concat layer

下载: 全尺寸图片幻灯片

图 11 添加内容损失函数前后增强效果对比

Figure 11. Comparison of enhancement effect before and after adding context loss function

下载: 全尺寸图片幻灯片

表 1 本文方法下图像增强客观评价结果

Table 1. Objective evaluation results of image enhancement of the proposed method

图像编号 PSNR SSIM
原始图像增强图像原始图像增强图像

T112 13.54 21.89 0.43 0.78
T118 14.55 18.65 0.13 0.74
T120 12.19 21.21 0.30 0.73

下载: 导出CSV

表 2 EM添加Concat层前后NIQE值对比

Table 2. Comparison of NIQE value before and after enhancement module adding Concat layer

图像编号 NIQE
无Concat层添加Concat层

T801 2.41 2.35
T822 3.78 3.59
T828 4.49 4.24
T841 2.83 2.58
平均值 3.38 3.19

下载: 导出CSV

参考文献(21)

[1]	钱鸣高,许家林,王家臣. 再论煤炭的科学开采[J]. 煤炭学报,2018,43(1):1-13. QIAN Minggao,XU Jialin,WANG Jiachen. Further on the sustainable mining of coal[J]. Journal of China Coal Society,2018,43(1):1-13.
[2]	谢斌红,袁帅,龚大立. 基于RDB−YOLOv4的煤矿井下有遮挡行人检测[J]. 计算机工程与应用,2022,58(5):200-207. XIE Binhong,YUAN Shuai,GONG Dali. Detection of blocked pedestrians based on RDB-YOLOv4 in coal mine[J]. Computer Engineering and Applications,2022,58(5):200-207.
[3]	李少荣. 基于改进直方图均衡化的红外图像增强技术的研究[J]. 工业控制计算机,2022,35(12):52-53,56. doi: 10.3969/j.issn.1001-182X.2022.12.019 LI Shaorong. Infrared image enhancement technology based on improved histogram equalization[J]. Industrial Control Computer,2022,35(12):52-53,56. doi: 10.3969/j.issn.1001-182X.2022.12.019
[4]	吕侃徽,张大兴. 基于自适应直方图均衡化耦合拉普拉斯变换的红外图像增强算法[J]. 光学技术,2021,47(6):747-753. doi: 10.13741/j.cnki.11-1879/o4.2021.06.018 LYU Kanhui,ZHANG Daxing. Infrared image enhancement algorithm based on adaptive histogram equalization coupled with Laplace transform[J]. Optical Technique,2021,47(6):747-753. doi: 10.13741/j.cnki.11-1879/o4.2021.06.018
[5]	杨微,姚冰莹,朱晓凤. 基于Retinex理论的低照度图像增强技术研究[J]. 现代计算机,2020(29):48-54. doi: 10.3969/j.issn.1007-1423.2020.29.008 YANG Wei,YAO Bingying,ZHU Xiaofeng. Reserch on low illumination image enhancement technology based on Retinex theory[J]. Modern Computer,2020(29):48-54. doi: 10.3969/j.issn.1007-1423.2020.29.008
[6]	陈超. 改进单尺度Retinex算法在图像增强中的应用[J]. 计算机应用与软件,2013,30(4):55-57. doi: 10.3969/j.issn.1000-386x.2013.04.016 CHEN Chao. Application of improved single scale Retinex algorithm in image enhancement[J]. Computer Applications and Software,2013,30(4):55-57. doi: 10.3969/j.issn.1000-386x.2013.04.016
[7]	赵晓霞,王汝琳,李雪艳. 基于多尺度Retinex的雾天降质图象增强算法[J]. 工矿自动化,2009,35(10):62-66. ZHAO Xiaoxia,WANG Rulin,LI Xueyan. Enhancement algorithm of fog-degraded image based on multiscale Retinex[J]. Industry and Mine Automation,2009,35(10):62-66.
[8]	WANG Shuhang,ZHENG Jin,HU Haimiao,et al. Naturalness preserved enhancement algorithm for non-uniform illumination images[J]. IEEE Transactions on Image Processing,2013,22(9):3538-3548. doi: 10.1109/TIP.2013.2261309
[9]	GUO Xiaojie,LI Yu,LING Haibin. LIME:low-light image enhancement via illumination map estimation[J]. IEEE Transactions on Image Processing,2016,26(2):982-993.
[10]	胡建平,郝梦云,杜影,等. 结构和纹理感知的Retinex融合红外与可见光图像[J]. 光学精密工程,2022,30(24):3225-3238. doi: 10.37188/OPE.20223024.3225 HU Jianping,HAO Mengyun,DU Ying,et al. Fusion of infrared and visible images via structure and texture-aware Retinex[J]. Optics and Precision Engineering,2022,30(24):3225-3238. doi: 10.37188/OPE.20223024.3225
[11]	DONG Xuan, WANG Guan, PANG Yi, et al. Fast efficient algorithm for enhancement of low lighting video[C]. IEEE International Conference on Multimedia and Expo, Barcelona, 2011: 1-6.
[12]	LORE K G,AKINTAYO A,SARKAR S. LLNet:a deep autoencoder approach to natural low-light image enhancement[J]. Pattern Recognition,2017,61:650-662. doi: 10.1016/j.patcog.2016.06.008
[13]	TAO Li, ZHU Chuang, XIANG Guoqing, et al. LLCNN: a convolutional neural network for low-light image enhancement[C]. IEEE Visual Communications and Image Processing, Petersburg, 2017: 10-13.
[14]	WANG Wenjing, WEI Chen, YANG Wenhan, et al. GLADNet: low-light enhancement network with global awareness[C]. 13th IEEE International Conference on Automatic Face & Gesture Recognition, Xi'an, 2018: 15-19.
[15]	WEI Chen, WANG Wenjing, YANG Wenhan, et al. Deep Retinex decomposition for low-light enhancement[C]. British Machine Vision Conference, Newcastle, 2018: 1-12.
[16]	王满利, 张航, 李佳悦, 等. 基于深度神经网络的矿井下低光照图像增强算法[J/OL]. 煤炭科学技术: 1-13[2023-04-20]. https://doi.org/10.13199/j.cnki.cst.2022-1626. WANG Manli, ZHANG Hang, LI Jiayue, et al. Deep neural network-based image enhancement algorithm for low-illumination images underground mines[J/OL]. Coal Science and Technology: 1-13[2023-04-20]. https://doi.org/10.13199/j.cnki.cst.2022-1626.
[17]	GUO Chunle, LI Chongyi, GUO Jichang, et al. Zero-reference deep curve estimation for low-light image enhancement[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, 2020: 13-19.
[18]	LYU Feifan, LU Feng, WU Jianhua, et al. MBLLEN: low-light image/video enhancement using CNNs[C]. British Machine Vision Conference, Newcastle, 2018: 220-233.
[19]	朱新山,姚思如,孙彪,等. 图像质量评价:融合视觉特性与结构相似性指标[J]. 哈尔滨工业大学学报,2018,50(5):121-128. ZHU Xinshan,YAO Siru,SUN Biao,et al. Image quality assessment:combining the characteristics of HVS and structural similarity index[J]. Journal of Harbin Institute of Technology,2018,50(5):121-128.
[20]	孙彦景,杨玉芬,刘东林,等. 基于内在生成机制的多尺度结构相似性图像质量评价[J]. 电子与信息学报,2016,38(1):127-134. SUN Yanjing,YANG Yufen,LIU Donglin,et al. Multiple-scale structural similarity image quality assessment based on internal generative mechanism[J]. Journal of Electronics & Information Technology,2016,38(1):127-134.
[21]	单月,刘段,万晓霞. 无参考图像质量评价研究现状与前景分析[J]. 包装工程,2022,43(13):296-304. doi: 10.19554/j.cnki.1001-3563.2022.13.037 SHAN Yue,LIU Duan,WAN Xiaoxia. Current research status and prospect of no-reference image quality assessment[J]. Packaging Engineering,2022,43(13):296-304. doi: 10.19554/j.cnki.1001-3563.2022.13.037