基于边界约束的煤矿井下尘雾图像去雾算法

曹虎晨; 姚善化; 王仲根

doi:10.13272/j.issn.1671-251x.2022010010

基于边界约束的煤矿井下尘雾图像去雾算法

安徽理工大学电气与信息工程学院, 安徽淮南　232001

基金项目: 安徽省自然科学基金面上项目(2108085MF200)。

详细信息

作者简介:
曹虎晨(1998—)，男，安徽蚌埠人，硕士研究生，研究方向为图像去雾，E-mail：2917443541@qq.com

中图分类号: TD714
计量
- 文章访问数: 328
- HTML全文浏览量: 63
- PDF下载量: 107
出版历程
- 收稿日期: 2022-01-05
- 修回日期: 2022-05-29
- 网络出版日期: 2022-05-04
- 刊出日期: 2022-06-29

Defogging algorithm of underground coal mine dust and fog image based on boundary constraint

School of Electrical and Information Engineering, Anhui University of Science and Technology, Huainan 232001, China

摘要

摘要: 现有煤矿井下图像去雾算法主要有基于图像增强的去雾算法、基于CNN的去雾算法和基于物理模型的去雾算法。前两者去雾效果差，易出现过曝光。基于物理模型的去雾算法依据大气散射模型对尘雾进行处理，但将基于暗通道的大气光值估计方法应用到煤矿井下环境中，选取的大气光值会较小，易造成图像过曝光、无法抑制点光源照射等问题。针对上述问题，将基于暗原色先验的图像去雾算法（He算法）与基于边界约束及上下文正则化的去雾算法（Meng算法）进行融合，提出了一种基于边界约束的煤矿井下尘雾图像去雾算法。首先，对输入的图像进行伽马校正，对校正后的图像进行颜色通道开运算处理，得到低分辨率的像素块，并从中选取最大亮度值作为煤矿井下的大气光值。其次，分别用He算法与Meng算法对伽马校正后的图像进行处理，对采用Meng算法得到的边界约束图进行引导滤波，得到更为清晰的边界约束图，并将Meng算法与He算法的粗透射率差值进行比较再融合。最后，对融合后的粗透射率进行上下文正则化得到细化透射率，根据得出的大气光值与细化后的透射率，通过大气散射模型得到去雾后的图像。仿真结果表明，基于边界约束的煤矿井下尘雾图像去雾算法没有出现过曝光等问题，且对浓雾图像的去雾效果更好，去雾后的图像也更明亮，颜色更加接近原图。采用峰值信噪比（PSNR）、结构相似性指标(SSIM)、特征相似性指标(FSIM）3种指标对去雾效果进行客观评价，结果表明，提出的算法在PSNR，SSIM，FSIM上相较于He算法平均提升了61.52%，36.51%，24.57%，相较于文献[9]算法平均提升了15.51%，19.27%，−0.30%，相较于Meng算法平均提升了18.93%，7.19%，1.21%，相较于文献[11]算法平均提升了18.29%，10.54%，1.19%，说明提出的算法在煤矿井下环境中去雾效果更好、图像更加明亮、细节信息保留更多。
- 煤矿井下图像 /
- 尘雾图像 /
- 去雾算法 /
- 伽马校正 /
- 暗通道先验 /
- 边界约束 /
- 引导滤波 /
- 透射率 /
- 上下文正则化 /
- 大气光值
Abstract: The existing defogging algorithms of underground coal mine images mainly include defogging algorithm based on image enhancement, defogging algorithm based on CNN and defogging algorithm based on physical model. The former two have poor defogging effect and are prone to over-exposure. The physical model-based defogging algorithm processes the dust and fog according to the atmospheric scattering model. However, the dark channel-based atmospheric light value estimation method is applied to the underground coal mine environment, and the selected atmospheric light value will be small. The problems of image overexposure, incapability of inhibiting point light source irradiation are easily caused. In order to solve the above problems, the image defogging algorithm based on dark primary color prior (He algorithm) and the defogging algorithm based on boundary constraint and context regularization (Meng algorithm) are fused. The defogging algorithm of underground coal mine dust and fog image based on boundary constraint is proposed. The method comprises the following steps. Firstly, Gamma correction is performed on the input image. The color channel opening operation processing is performed on the corrected image to obtain the low-resolution pixel block. The maximum brightness value is selected from the low-resolution pixel block as the underground atmospheric light value of the coal mine. Secondly, the Gamma-corrected image is processed by He algorithm and Meng algorithm respectively. The boundary constraint map obtained by Meng algorithm is filtered to obtain a clearer boundary constraint map. And the rough transmittance difference of Meng algorithm and He algorithm is compared and then fused. Finally, the contextual regularization is performed on fused rough transmittance to obtain the refined transmittance. The obtained atmospheric light value and the refined transmittance are used to obtain the defogged image through the atmospheric scattering model. The simulation results show that the proposed defogging algorithm of underground coal mine dust and fog image based on boundary constraint has no problems such as overexposure. The defogging effect is better, the defogged image is brighter and the color is closer to the original image. The peak signal-to-noise ratio (PSNR), structural similarity index (SSIM) and feature similarity index (FSIM) are used to objectively evaluate the defogging effect of the proposed algorithm. The results show that the proposed algorithm has an average improvement of 61.52%, 36.51% and 24.57% in PSNR, SSIM and FSIM compared with the He algorithm. Compared with the algorithm in literature [9], the proposed algorithm has increased by 15.51%, 19.27% and −0.30% on average. Compared with Meng algorithm, the proposed algorithm has increased by 18.93%, 7.19% and 1.21% on average. Compared with the algorithm in literature [11], the proposed algorithm has increased by 18.29%, 10.54% and 1.19% on average. It shows that the proposed algorithm has better defogging effect, brighter image and more details in the underground environment of coal mine.
- coal mine underground image /
- dust and fog image /
- defogging algorithm /
- Gamma correction /
- dark channel prior /
- boundary constraint /
- guided filtering /
- transmittance /
- contextual regularization /
- atmospheric light value

HTML全文

图 1 去雾算法流程

Figure 1. Flow chart of defogging algorithm

下载: 全尺寸图片幻灯片

图 2 伽马校正结果

Figure 2. Results of Gamma correction

下载: 全尺寸图片幻灯片

图 3 大气光值估计方法对比

Figure 3. Comparison of atmospheric light estimation algorithms

下载: 全尺寸图片幻灯片

图 4 引导滤波结果

Figure 4. Results of guiding filtering

下载: 全尺寸图片幻灯片

图 5 融合结果

Figure 5. Result of fusion

下载: 全尺寸图片幻灯片

图 6 去雾效果对比

Figure 6. Comparison of defogging effect

下载: 全尺寸图片幻灯片

表 1 不同算法去雾图像指标比较

Table 1 Indicators comparison of defogging images processed by different algorithms

图像	算法	评价指标
图像	算法	PSNR	SSIM	FSIM
图像1	He算法	18.898 982 45	0.815 1	0.917 9
	文献[9]算法	14.077 211 43	0.763 9	0.954 0
	Meng算法	13.873 425 61	0.753 2	0.935 5
	文献[11]算法	13.996 310 17	0.735 8	0.938 0
	本文算法	16.683 743 23	0.856 3	0.949 9
图像2	He算法	18.202 339 15	0.854 2	0.878 0
	文献[9]算法	18.639 958 14	0.895 8	0.950 6
	Meng算法	16.317 298 54	0.871 3	0.935 9
	文献[11]算法	16.454 639 11	0.847 4	0.934 7
	本文算法	19.681 150 10	0.812 0	0.947 1
图像3	He算法	15.751 470 56	0.761 9	0.854 9
	文献[9]算法	15.790 684 38	0.738 7	0.957 5
	Meng算法	14.872 303 84	0.721 8	0.928 1
	文献[11]算法	14.984 407 73	0.672 4	0.921 0
	本文算法	17.286 198 78	0.821 6	0.953 0
图像4	He算法	11.509 785 38	0.617 0	0.756 3
	文献[9]算法	14.239 536 23	0.488 8	0.940 7
	Meng算法	16.248 842 47	0.784 1	0.949 4
	文献[11]算法	16.348 279 20	0.780 3	0.954 3
	本文算法	19.326 245 59	0.869 4	0.961 3
图像5	He算法	6.726 802 218	0.357 4	0.580 8
	文献[9]算法	18.191 280 39	0.783 5	0.955 4
	Meng算法	17.127 039 24	0.783 5	0.935 9
	文献[11]算法	17.015 085 49	0.753 7	0.936 2
	本文算法	20.310 984 51	0.852 2	0.946 6
图像6	He算法	13.360 759 69	0.706 7	0.710 5
	文献[9]算法	21.630 907 30	0.809 6	0.973 6
	Meng算法	20.386 995 18	0.906 2	0.961 1
	文献[11]算法	20.556 107 84	0.892 4	0.963 0
	本文算法	24.251 961 61	0.931 3	0.955 8

下载: 导出CSV

参考文献(13)

[1]	LIU Zhigang,CAO Anye,GUO Xiaosheng,et al. Deep-hole water injection technology of strong impact tendency coal seam—a case study in Tangkou Coal Mine[J]. Arabian Journal of Geosciences,2018,11(2):1-9.
[2]	肖军良. 辛置煤矿2-208工作面喷雾降尘技术研究与应用[J]. 煤矿现代化,2021,30(6):46-48. DOI: 10.3969/j.issn.1009-0797.2021.06.014 XIAO Junliang. Research and application of spray dust suppression technology in 2-208 working face of Xinzhi Coal Mine[J]. Coal Mine Modernization,2021,30(6):46-48. DOI: 10.3969/j.issn.1009-0797.2021.06.014
[3]	王道累,张天宇. 图像去雾算法的综述及分析[J]. 图学学报,2020,41(6):861-870. WANG Daolei,ZHANG Tianyu. Review and analysis of image defogging algorithm[J]. Journal of Graphics,2020,41(6):861-870.
[4]	张立亚, 郝博南, 孟庆勇, 等. 基于HSV空间改进融合Retinex算法的井下图像增强方法[J]. 煤炭学报, 2020, 45（增刊1）: 532-540. ZHANG Liya, HAO Bonan, MENG Qingyong, et al. Method of image enhancement in coal mine based on improved Retinex fusion algorithm in HSV space[J]. Journal of China Coal Society, 2020, 45（S1）: 532-540.
[5]	龚云,杨庞彬,颉昕宇. 结合同态滤波与直方图均衡化的井下图像匹配算法[J]. 工矿自动化,2021,47(10):37-41,61. GONG Yun,YANG Pangbin,JIE Xinyu. Underground image matching algorithm combining homomorphic filtering and histogram equalization[J]. Industry and Mine Automation,2021,47(10):37-41,61.
[6]	智宁,毛善君,李梅,等. 基于深度融合网络的煤矿图像尘雾清晰化算法[J]. 煤炭学报,2019,44(2):655-666. ZHI Ning,MAO Shanjun,LI Mei,et al. Coal mine image dust and fog clearing algorithm based on deep fusion network[J]. Journal of China Coal Society,2019,44(2):655-666.
[7]	HE Kaiming,SUN Jian,TANG Xiao'ou. Single image Haze removal using dark channel prior[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2011,33(12):2341-2353. DOI: 10.1109/TPAMI.2010.168
[8]	刘晓文,仲亚丽,袁莎莎,等. 基于暗原色先验的煤矿井下退化图像复原算法[J]. 煤炭科学技术,2012,40(6):77-80. LIU Xiaowen,ZHONG Yali,YUAN Shasha,et al. Restoration algorithms of degradation image in underground mine based on dark channel prior[J]. Coal Science and Technology,2012,40(6):77-80.
[9]	杜明本,陈立潮,潘理虎. 基于暗原色理论和自适应双边滤波的煤矿尘雾图像增强算法[J]. 计算机应用,2015,35(5):1435-1438,1448. DOI: 10.11772/j.issn.1001-9081.2015.05.1435 DU Mingben,CHEN Lichao,PAN Lihu. Enhancement algorithm for fog and dust images in coal mine based on dark channel prior theory and bilateral adaptive filter[J]. Journal of Computer Applications,2015,35(5):1435-1438,1448. DOI: 10.11772/j.issn.1001-9081.2015.05.1435
[10]	MENG Gaofeng, WANG Ying, DUAN Jiangyong, et al. Efficient image dehazing with boundary constraint and contextual regularization[C]//2013 IEEE International Conference on Computer Vision, Sydney, 2013: 617-624.
[11]	杨红,崔艳. 基于开运算暗通道和优化边界约束的图像去雾算法[J]. 光子学报,2018,47(6):244-250. YANG Hong,CUI Yan. Image defogging algorithm based on opening dark channel and improved boundary constraint[J]. Acta Photonica Sinica,2018,47(6):244-250.
[12]	NARASIMHAN S G, NAYAR S K. Chromatic framework for vision in bad weather[C]//IEEE Computer Society Conference on Computer Vision & Pattern Recognition, Hilton Head Island, 2000: 598-605
[13]	HE Kaiming,SUN Jian,TANG Xiao'ou. Guided image filtering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(6):1397-1409. DOI: 10.1109/TPAMI.2012.213

施引文献(9)

期刊类型引用(7)

1.	刘宇超，高妍，张红娟. 矿用带式输送机多目标参数协同优化控制储能方法. 煤炭技术. 2024(04): 246-249 . 百度学术
2.	罗家林，陈超，黄梅，辛迪熙，吕洪章. 弱电网矿区下构网型储能控制技术研究. 高压电器. 2023(07): 95-103 . 百度学术
3.	王磊，王凯. 长距离掘进工作面局部通风智能联动调控研究. 工矿自动化. 2023(09): 55-63 . 本站查看
4.	李春，陈国栋. 储能逆变器测试平台设计和实现. 信息记录材料. 2022(08): 168-171 . 百度学术
5.	程晓之，王凯，郝海清，陈瑞鼎，吴建宾. 矿井局部通风智能调控系统及关键技术研究. 工矿自动化. 2021(09): 18-24 . 本站查看
6.	文小珲. 基于模糊优化算法的电力应急电源设备自动化配置方法. 能源与环保. 2021(12): 202-206+211 . 百度学术
7.	李正山. 矿井通风机供电系统设计探究. 能源与节能. 2020(08): 93-94 . 百度学术