随着煤矿智能化技术的不断发展，煤矿安全化生产受到煤矿企业的广泛关注。基于煤矿监控系统全天候、实时性、连续性的特点，自动采集图像实现井下目标的精准检测跟踪成为了当前的研究热点。由于井下环境的复杂性，造成获取的图像存在光照不足、照度低等干扰因素，难以满足煤矿智能化管控平台的要求，影响后续目标检测及识别分析。为提供井下真实场景信息，保障人员生命安全，本文针对煤矿井下图像分别从图像增强和目标检测两方面进行研究，具体研究内容如下：

（1）针对井下图像存在照度低、光照不均的问题，本文研究基于中心环绕的Retinex算法以及图像融合增强方法，提出了煤矿井下低照度图像增强策略。首先采用卷积神经网络（Convolutional Neural Network，CNN）将图像分解为光照分量和反射分量，得到初始增强图像；其次，引入相机响应模型，根据“图像熵最大化”原则生成虚拟过曝光图像；最后结合原始图像，采用图像块分解进行融合重构得到增强图像。实验结果表明，相较于主流的增强算法，本文的增强策略在主观上更接近人眼对场景的理解，在客观指标上整体表现良好，表明本方法对于煤矿井下图像增强具有较好的适用性。

（2）针对传统SSD（Single Shot multibox Detector）目标检测模型参数量过大、检测效率较低的问题，本文引入倒置残差结构（Inverted Residual block）和线性瓶颈层（Linear Bottlenecks）替代常规的标准卷积，搭建以MobileNetV2为主干特征提取网络的轻量化检测模型；为进一步提升检测精度，引入广义交并比损失(GIOU Loss)作为位置回归损失函数，以提高目标位置回归框定位的准确率。实验结果表明，本文检测模型平均精确度达到87.83%，与传统SSD检测算法相比，平均精确度提高了0.89个百分点，模型大小减少了52.44MB，检测速度提升了1.53帧/每秒。

With the continuous development of coal mine intelligent technology, coal mine safety production is widely concerned by coal mine enterprises. Based on the characteristics of total-weather, real-time, continuity, automatic acquisition of image implementation, and the precision detection tracking of the downhole targets is a current research hotspot. Due to the complexity of the downhole environment, the acquired image has an insufficient illuminance, low illuminance and other interference factors, it is difficult to meet the requirements of the coal mine intelligent management platform, affect subsequent goal detection and identification analysis. In order to provide the real scene information in the downhole, the safety of the personnel is safe. This paper is studied from both image enhancement and target detection in the coal mine underground image. The specific research content is as follows:

(1) Aiming at the problem of low illumination and uneven light image, this paper studies the centrally surround Retinex algorithm and image fusion enhancement method, proposes the low illumination image enhancement strategy under the coal mine well. Firstly, the Convolutional Neural Network, CNN is used to decompose the image into the light component and the reflective component to obtain an initial enhanced image; secondly, the introduction camera response model is introduced to generate a virtual exposure image according to the "image entropy maximization" principle; final In conjunction with the original image, the fusion reconstruction of the image block decomposition is used to enhance the image. The experimental results show that compared to mainstream enhancement algorithms, this paper's enhancement strategy is subjectively more close to the human eye to the scene's eye, the overall performance is good in objective indicators, indicating that the method has better applicability to the image of the coal mine.

(2) For the problem of traditional SSD (Single Shot Multibox Detector) target detection model parameters and low detection efficiency, this article introduces inverted residual blocks and linear bottlenecks to replace conventional standard convolution. Establish a lightweight detection model with MobileNetV2 as the main characteristics to extract the network; in order to further improve the detection accuracy, introduce a broader and giou Loss to the loss function to improve the accuracy of the positioning of the target position regression box. The experimental results show that compared with the traditional SSD detection algorithm, the average accuracy of the detection model in this paper increased by 0.89%, the recall rate increased by 0.51%, the model size decreased by 52.44MB, the detection speed increased by 1.53 frames/per second.

[1] 王国法, 杜毅博, 任怀伟, 等. 智能化煤矿顶层设计研究与实践[J]. 煤炭学报, 2020, 45(06): 1909-1924.

[2] 丁全利,胡容波. 中国矿产资源报告(2021)[N]. 中国自然资源报,2021-10-22(001).

[3] 樊大磊, 李富兵, 王宗礼, 等. 碳达峰、碳中和目标下中国能源矿产发展现状及前景展望[J]. 中国矿业, 2021, 30(6): 1-8.

[4] 国家安全生产监督管理局. 国家矿山安监局要闻[EB/OL]. https: //www. chinamine-safety. gov. cn/xw/mkaqjcxw/202203/t20220306_409151. shtml 2022. 3. 6.

[5] 能源技术革命创新行动计划(2016-2030年)[J]. 电力与能源, 2016, 37(3): 1.

[6] 王勇. 煤矿救灾机器人井下可视导航技术研究[D]. 中国矿业大学, 2018.

[7] 彭苏萍. 我国煤矿安全高效开采地质保障系统研究现状及展望[J]. 煤炭学报, 2020, 45(7): 2331-2345.

[8] 张俊文, 杨虹霞. 2005—2019年我国煤矿重大及以上事故统计分析及安全生产对策研究[J]. 煤矿安全, 2021, 52(12): 261-264.

[9] 程程. 国家矿山安全监察局举办首次新闻发布会[J]. 中国安全生产, 2021, 16(01): 11-13.

[10] 孙继平. 煤矿安全生产理念研究[J]. 煤炭学报, 2011, 36(2): 313-316.

[11] 彭担任, 李世明. 提高煤矿生产安全管理的对策[J]. 煤矿安全, 2005, 36(10): 63-65.

[12] 刘备战, 赵洪辉, 周李兵. 面向无人驾驶的井下行人检测方法[J]. 工矿自动化, 2021, 47(9): 113-117.

[13] 李晓建. 矿井人员目标检测与跟踪算法的研究与实现[D]. 山东科技大学, 2020.

[14] 王洪栋, 郭伟东, 朱美强, 等. 一种煤矿井下低照度图像增强算法[J]. 工矿自动化, 2019, 45(11): 81-85.

[15] 张莉敏, 祝明慧. 煤矿提升安全状态监测预警系统设计[J]. 煤炭技术, 2017, 36(5): 251-253.

[16] 毛善君. “高科技煤矿”信息化建设的战略思考及关键技术[J]. 煤炭学报, 2014, 39(8): 1572-1583.

[17] 韩江洪, 卫星, 陆阳, 等. 煤矿井下机车无人驾驶系统关键技术[J]. 煤炭学报, 2020, 45(6): 2104-2115.

[18] 孙继平. 煤矿井下人员位置监测技术与系统[J]. 煤炭科学技术, 2010, 38(11): 1-5.

[19] 董观利, 宋春林. 基于视频的矿井行人越界检测系统[J]. 工矿自动化, 2017, 43(2): 29-34.

[20] 杨铮. 基于视频矿井下的人员计数算法研究[D]. 武汉理工大学, 2013.

[21] 王浩, 张叶, 沈宏海, 等. 图像增强算法综述[J]. 中国光学, 2017, 10(4): 438-448.

[22] Gonzalez R C, Woods R E. Digital image processing[J]. Prentice Hall International, 2008, 28(4): 484-486.

[23] 张岩, 崔晓萌. 基于灰度变换的图像增强实现[J]. 包装工程, 2010, 31(19): 95-98.

[24] 秦钟, 杨建国, 王海默, 等. 基于Retinex理论的低照度下输电线路图像增强方法及应用[J]. 电力系统保护与控制, 2021, 49(3): 150-157.

[25] 曾鹏鑫, 么健石, 陈鹏, 等. 基于小波变换的图像增强算法[J]. 东北大学学报（自然科学版）, 2005, 26(6): 527-530.

[26] 余春艳, 徐小丹, 林晖翔, 等. 应用雾天退化模型的低照度图像增强[J]. 中国图象图形学报, 2017, 22(9): 1194-1205.

[27] Kim J Y, Kim L S, Hwang S H. An advanced contrast enhancement using partially overlapped sub-block histogram equalization[C]// 2000 IEEE International Symposium on Circuits and Systems (ISCAS). IEEE, 2002

[28] 武英. 保持图像亮度的改进双直方图均衡算法[J]. 计算机应用, 2010, 30(6): 1632-1634.

[29] Hui D I, Qi-Feng Y U, Zhang X H. An algorithm for infrared image enhancement based on gray scale transform[J]. Journal of Applied Optics, 2006, 27(1): 12-14.

[30] 廖斌, 刘鸳鸳. 基于多尺度灰度变换的图像增强研究[J]. 量子电子学报, 2015, 32(5): 550-554.

[31] Jobson, Daniel, J, et al. Properties and performance of a center/surround retinex[J]. IEEE Transactions on Image Processing, 1997.

[32] Rahman Z, Jobson D J, Woodell G. Multiscale retinex for color image enhancement[J]. Proc. intl Conf. on Image Processing, 1996.

[33] Jobson D J, Rahman Z, Woodell G A. A multiscale retinex for bridging the gap between color images and the human observation of scenes[J]. IEEE Transactions on Image Processing, 2002, 6(7): 965-976.

[34] Tingting S, Cheolkon J. Readability enhancement of low light images based on dual-tree complex wavelet transform[C]// IEEE International Conference on Acoustics. IEEE, 2016.

[35] Li C, Fan T, Xiao M, et al. An improved image defogging method based on dark channel prior[C]// 2017 2nd International Conference on Image, Vision and Computing (ICIVC). 2017.

[36] 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.

[37] Li C, J Guo, Porikli F, et al. LightenNet: A Convolutional Neural Network for weakly illuminated image enhancement[J]. Pattern recognition letters, 2018, 104(MAR. 1): 15-22.

[38] 黄鐄, 陶海军, 王海峰. 条件生成对抗网络的低照度图像增强方法[J]. 中国图象图形学报, 2019, 24(12): 2149-2158.

[39] Guo C, Li C, J Guo, et al. Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement[J]. IEEE, 2020.

[40] 江泽涛, 覃露露. 一种基于U-Net生成对抗网络的低照度图像增强方法[J]. 电子学报, 2020, 48(2): 258-264. DOI: 10. 3969/j. issn. 0372-2112. 2020. 02. 005.

[41] 李新锋. 煤矿视频监控图像增强方法的研究[D]. 黑龙江科技学院, 2010.

[42] Zhao Y Q, Wang Z C, Xi-Hui F U. Research on Image Enhancement Based on Coal Mine Video Monitoring System[J]. Coal Mine Machinery, 2012.

[43] Shang C, Ma H, Qi Z, et al. Coal Mine Machine Vision Image Enhancement Technology Research[J]. Computer Measurement & Control, 2013.

[44] 吕建中. 基于图像的井下人员检测算法研究[D]. 重庆大学, 2015.

[45] 雷耀花. 煤矿井下视频图像增强与人员检测技术研究[D]. 太原科技大学, 2015

[46] Yu C, Liu X L, Deng L J. An Enhancement Method for Non-uniform Illuminationin Coal Mine[C]// International Symposium on Computer. IEEE, 2016.

[47] 王诚聪, 刘亚静. 矿井复杂环境视频监控图像增强算法研究[J]. 煤炭工程, 2021, 53(4): 147-151.

[48] Viola P A, Jones M J. Rapid Object Detection using a Boosted Cascade of Simple Features[C]// Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society Conference on. IEEE, 2001.

[49] Dalal N, Triggs B. Histograms of Oriented Gradients for Human Detection[C]// IEEE Computer Society Conference on Computer Vision & Pattern Recognition. IEEE, 2005.

[50] Felzenszwalb, Pedro, F, et al. Object Detection with Discriminatively Trained Part-Based Models. [J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2010, 32(9): 1627-1645.

[51] He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2015, 37(9): 1904-1916.

[52] Girshick R. Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2015: 1440-1448.

[53] Ren S, He K, Girshick R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(6): 1137-1149.

[54] Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788.

[55] Liu W, Anguelov D, Erhan D, et al. Ssd: Single shot multibox detector[C]//European conference on computer vision. Springer, Cham, 2016: 21-37.

[56] Fu C Y, Liu W, Ranga A, et al. DSSD: Deconvolutional Single Shot Detector[J]. 2017.

[57] Jeong J, Park H, Kwak N. Enhancement of SSD by concatenating feature maps for object detection[J]. 2017.

[58] Dickens J S, Wyk M, Green J J. Pedestrian detection for underground mine vehicles using thermal images[C]// Africon. IEEE, 2011.

[59] 朱光泽. 基于目标检测与跟踪算法的煤矿井下视频监控系统研究[D]. 辽宁工程技术大学, 2014.

[60] 雷耀花. 煤矿井下视频图像增强与人员检测技术研究[D]. 太原科技大学, 2015.

[61] 郑嘉祺. 基于DCNN的井下行人检测系统的研究与设计[D]. 西安科技大学, 2017.

[62] 李伟山, 卫晨, 王琳. 改进的Faster RCNN煤矿井下行人检测算法[J]. 计算机工程与应用, 2019, 55(4): 200-207.

[63] 李珊. 基于视频的矿井下人员检测方法研究[D]. 武汉理工大学, 2018.

[64] 石永恒, 杨超宇. 基于深度学习的矿井下作业人员安全帽佩戴检测[J]. 绥化学院学报, 2021, 41(09): 148-152.

[65] 魏东. 采煤机工作空间人员检测与预警关键技术研究[D]. 中国矿业大学, 2021.

[66] DE Rumelhart, Hinton G, Williams R J. Learning Representations by Back Propagating Errors[J]. Nature, 1986, 323(6088): 533-536.

[67] Lecun Y, Bottou L. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324

[68] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[J]. Advances in neural information processing systems, 2012, 25: 1097-1105.

[69] Simonyan K, Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition[J]. Computer Science, 2014.

[70] K. He, X. Zhang, S. Ren and J. Sun, "Deep Residual Learning for Image Recognition, " 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770-778.

[71] Szegedy C, Liu W, Jia Y, et al. Going Deeper with Convolutions[J]. IEEE Computer Society, 2014.

[72] Land E H. The Retinex Theory of Color Vision[J]. Scientific American, 1978, 237(6): 108-128.

[73] Goshtasby A A. Fusion of multi-exposure images[J]. Image and Vision Computing, 2005, 23(6): 611-618.

[74] Bachoo A K. Real-time exposure fusion on a mobile computer[J]. Prasa, 2009.

[75] Grossberg M D, Nayar S K. What is the space of camera response functions?[C]//2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings. IEEE, 2003, 2: II-602.

[76] Mann, Steve. Comparametric Equations with Practical Applications in QuantigraphicImage Processing. [J]. IEEE Transactions on Image Processing, 2000.

[77] Guo X, Yu L, Ling H. LIME: Low-light Image Enhancement via Illumination Map Estimation[J]. IEEE Transactions on Image Processing, 2016, PP(99): 1-1.

[78] Ying Z, Ge L, Wen G. A Bio-Inspired Multi-Exposure Fusion Framework for Low-light Image Enhancement[J]. 2017.

[79] Wei C, Wang W, Yang W, et al. Deep Retinex Decomposition for Low-Light Enhancement[J]. 2018.

[80] 何南南, 解凯, 李桐, 叶宇姗. 图像质量评价综述[J]. 北京印刷学院学报, 2017, 25(02): 47-50. 2017. 02. 012.

[81] Wang K, Wang H, Li Y, et al. Quantitative performance evaluation for dehazing algorithms on synthetic outdoor hazy images[J]. IEEE Access, 2018, 6: 20481-20496.