煤炭资源作为我国的主体能源，对我国的经济发展有着极其重要的作用。在煤矿生产过程中，不可避免的会伴生固体废弃物煤矸石，尤其是采用综放开采工艺，在煤炭资源高效生产的同时也产生了大量的煤矸石，煤矸石不仅会影响原煤质量，增加煤炭生产成本，而且大量堆积会污染环境。因此，精准决策控制放煤，避免过放现象，减少煤矸石的产生是实现综放智能开采亟待解决的关键科学问题之一。随着计算机视觉技术的快速发展，基于机器视觉技术的智能化设备装置被广泛应用于工业生产活动。论文以精准辨识煤矸和指导放煤决策为目标，以陕西华彬煤业股份有限公司蒋家河煤矿1405工作面为研究背景，搭建基于机器视觉的暗湿工况下煤矸热敏识别平台，优化多目标实时识别检测算法，采集实验数据并提取煤矸运移态势热敏影像特征，提出一种混矸率测算方法。论文的主要研究内容和开展工作情况如下：

厘清了当前放煤控制与决策出现的实际影响因素，构建了运移煤矸热敏影像采集实验平台，采集完善了煤矸热敏影像数据库，通过暗通道去雾和CLAHE算法等预处理手段有效提升了监测效果，解决了暗湿工况以及粉尘环境影响煤矸识别准确率的问题；提出了一种基于SA-YOLOv7+Deep Sort的运移态势煤矸热敏影像识别算法，将注意力SA模块插入YOLOv7模型主干网络以提高对小目标煤矸的检测能力，引入部分卷积Pconv优化ELAN模块以提高模型对于煤矸影像关键特征的提取能力，改进损失函数提升模型泛化能力，改进后SA-YOLOv7模型的mAP和F-Measure值相较于原始模型分别提高了8.8%和10.2%；引入三元组损失优化Deep Sort目标跟踪算法表观特征网络性能，将原始的特征提取网络替换为Resnet18网络结构，优化了动态监测跟踪过程的准确率和速度，煤矸目标跟踪时间缩短至65 ms内，满足运移煤矸实时监测的实时性要求；采用扫描电子显微镜等技术手段探究煤和矸石微观结构差异性，揭示了运移态势煤矸热敏变化特征规律，暗湿工况下的煤矸温度变化均经历快速降低阶段、缓慢降低阶段、平稳波动阶段等三个阶段，且煤样的温度降低持续时间更长，温度变化幅度更大；基于实验过程中获取的煤矸热敏数据，提取热敏变化特征并提出了一种混矸率测算方法，可以较好的反映原煤运移过程中的真实混矸率。

论文提出的基于机器视觉的煤矸快速运移态势热敏影像识别技术等研究成果可有效提升实际工况下煤矸识别的精度，实现了运移态势煤矸混矸率准确测算，为精准辨识煤矸和指导放煤决策提供了技术支撑。

As China's main energy source, coal resources play an extremely important role in my country's economic development. In the process of coal mining, solid waste gangue is inevitably produced, especially when the fully mechanized caving mining process is adopted. While coal resources are efficiently produced, a large amount of gangue is also produced. Gangue not only affects the quality of raw coal and increases the cost of coal production, but also pollutes the environment when it accumulates in large quantities. Therefore, accurate decision-making to control coal release, avoid over-release, and reduce the generation of gangue are one of the key scientific issues that need to be solved in order to achieve fully mechanized caving intelligent mining. With the rapid development of computer vision technology, intelligent equipment and devices based on machine vision technology have been widely used in industrial production activities. With the goal of accurately identifying gangue and guiding coal release decisions, this paper takes the 1405 working face of Jiangjiahe Coal Mine of Shaanxi Huabin Coal Industry Co., Ltd. as the research background, builds a coal gangue thermal recognition platform under dark and wet conditions based on machine vision, optimizes the multi-target real-time recognition and detection algorithm, collects experimental data and extracts the thermal image features of coal gangue migration status, and proposes a mixed gangue rate calculation method. The main research content and work progress of the paper are as follows:

The actual influencing factors of current coal release control and decision-making were clarified, an experimental platform for the acquisition of thermal images of coal gangue in transport was built, and a coal gangue thermal image database was collected and improved. The monitoring effect was effectively improved through pre-processing methods such as dark channel defogging and CLAHE algorithm, and the problem that dark and wet conditions and dusty environment affect the accuracy of coal gangue recognition was solved; a coal gangue thermal image recognition algorithm for transport situation based on SA-YOLOv7+Deep Sort was proposed, and the attention SA module was inserted into the YOLOv7 model backbone network to improve the detection ability of small target coal gangue. The partial convolution Pconv was introduced to optimize the ELAN module to improve the model's ability to extract key features of coal gangue images, and the loss function was improved to improve the generalization ability of the model. The mAP and F-Measure values ​​of the improved SA-YOLOv7 model were increased by 8.8% and 10.2% respectively compared with the original model; the triple loss was introduced to optimize Deep Sort. The performance of the apparent feature network of the Sort target tracking algorithm was improved, and the original feature extraction network was replaced with the Resnet18 network structure, which optimized the accuracy and speed of the dynamic monitoring and tracking process. The target tracking time of coal gangue was shortened to within 65 ms, which met the real-time requirements of real-time monitoring of transported coal gangue. Scanning electron microscopy and other technical means were used to explore the differences in the microstructures of coal and gangue, revealing the characteristics of thermal sensitivity changes in the transport status of coal gangue. The temperature changes of coal gangue under dark and wet conditions all experienced three stages: rapid decrease stage, slow decrease stage, and stable fluctuation stage. The temperature decrease of coal samples lasted longer and the temperature change amplitude was larger. Based on the thermal sensitive data of coal gangue obtained during the experiment, the thermal sensitivity change characteristics were extracted and a mixed gangue rate calculation method was proposed, which can better reflect the actual mixed gangue rate in the process of raw coal transportation.

The research results proposed in the paper, such as the thermal image recognition technology for rapid coal gangue migration based on machine vision, can effectively improve the accuracy of coal gangue recognition under actual working conditions, realize the accurate calculation of coal gangue mixed ratio in migration situation, and provide technical support for accurate identification of coal gangue and guidance of coal release decision-making.

[1] 谢和平,吴立新,郑德志.2025年中国能源消费及煤炭需求预测[J].煤炭学报,2019,44 (7):1949-1960.

[2] 谢和平,任世华,谢亚辰,等.碳中和目标下煤炭行业发展机遇[J].煤炭学报,2021,46(7): 2197-2211.

[3] 王家臣.我国放顶煤开采的工程实践与理论进展[J].煤炭学报,2018,43(1):43-51.

[4] 王国法,刘峰,孟祥军,等.煤矿智能化(初级阶段)研究与实践[J].煤炭科学技术,2019(8): 1-36.

[5] 张映红.关于能源结构转型若干问题的思考及建议[J].国际石油经济,2021,29(2):1-15

[6] 秦容军.我国煤炭开采现状及政策研究[J].煤炭经济研究,2019,39(1):57-61.

[7] 孟宪锐,吴昊天,王国斌.我国厚煤层采煤技术的发展及采煤方法的选择[J].煤炭工程,2014,46(10):43-47.

[8] 王家臣.我国放顶煤开采的工程实践与理论进展[J].煤炭学报,2018,43(1):43-51.

[9] 赵学雷,徐向东,李会义.王庄煤矿大采高自动化综放工作面技术研究[J].煤矿机械, 2008,29(4):160-161.

[10] 王家臣,仲淑姮.我国厚煤层开采技术现状及需要解决的关键问题[J].中国科技论文, 2008,3(11):829-834.

[11] 孟贤正,夏永军,程国强,等.放顶煤开采顶煤初次垮落动力学分析[J].煤炭学报, 2008,33(8):841-844.

[12] 李建民,章之燕.开滦矿区复杂煤层综合机械化放顶煤开采技术[J].煤炭学报,2010(11): 1794-1799.

[13] 田妍,田丰.放顶煤开采过程煤矸识别技术发展现状及前景[J].煤炭工程,2018, 50(10):142-145.

[14] 黄福昌.兖州矿区综放开采技术[J].煤炭学报,2010,35(11):1778-1782.

[15] 于斌,徐刚,黄志增,等.特厚煤层智能化综放开采理论与关键技术架构[J].煤炭学报,2019,44(1):42-53.

[16] 刘长友,张宁波,郭凤岐,等.特厚煤层综放煤-矸-岩放落流动的时序规律及识别方法[J].煤炭学报,2022,47(1):137-151.

[17] 李灿华,向晓东,刘思,等.煤矸石环境危害性及其资源化利用[J].武钢技术,2016,54(2): 58-62.

[18] MCCOY J T, AURET L. Machine learning applications in minerals processing: A review[J]. Minerals Engineering, 2019,132:95-109.

[19] WANG W, ZHANG C. Separating coal and gangue using three-dimensional laser scanning [J]. International Journal of Mineral Processing, 2017,169:79-84.

[20] 李曼,段雍,曹现刚,等.煤矸分选机器人图像识别方法和系统[J].煤炭学报,2020,45(10): 3636-3644.

[21] 刘峰,曹文君,张建明,等.我国煤炭工业科技创新进展及“十四五”发展方向[J].煤炭学报,2021,46(1):1-15.

[22] 张吉雄,屠世浩,曹亦俊,等.煤矿井下煤矸智能分选与充填技术及工程应用[J].中国矿业大学学报,2021,50(3):417-430.

[23] Shakoor M H, Boostani R. Extended mapping local binary pattern operator for texture classification[J]. International Journal of Pattern Recognition and Artificial Intelligence, 2017,31(6):1750019.

[24] Fernández-Robles L, Azzopardi G, Alegre E, et al. Ma-chine-vision-based identification of broken inserts in edge-profile milling heads[J]. Robotics and Computer-Integrated Manufacturing, 2017(44):276-283.

[25] Tsai D, Hsieh Y. Machine vision-based positioning and inspection using expectation–maximization technique[J]. IEEE Transactions on Instrumentation and Measurement, 2017, 66(11):2858-2868.

[26] Li B. Research on geometric dimension measurement system of shaft parts based on machine vision[J]. EURASIP Journal on Image and Video Processing, 2018,2018(1):101.

[27] Sture, ye E R, Skavhaug A, et al. A 3D machine vision system for quality grading of Atlantic salmon[J]. Computers and Electronics in Agriculture, 2016(123):142-148.

[28] Omid M, Soltani M, Dehrouyeh M H, et al. An expert egg grading system based on machine vision and artificial intelligence techniques[J]. Journal of food engineering, 2013,118(1): 70-77.

[29] Momin M A, Yamamoto K, Miyamoto M, et al. Machine vision based soybean quality evaluation[J]. Computers and Electronics in Agriculture, 2017,140:452-460.

[30] Habib M T, Majumder A, Jakaria A Z M, et al. Machine vision based papaya disease recognition[J]. Journal of King Saud University-Computer and Information Sciences, 2020, 32(3): 300-309.

[31] Yu B, Zhang W, Cai Y. A lane departure warning system based on machine vision[C]//2008 IEEE Pacific-Asia Workshop on Computational Intelligence and Industrial Application. IEEE, 2008,1:197-201.

[32] 肖剑,杨钧宇,徐紫琪,等.基于机器视觉的螺纹钢丝头检测方法[J].制造业自动化,2022, 44(8):14-19.

[33] 刘科,钟志成,曹静.基于机器视觉的激光焊接焊缝智能检测研究[J].激光杂志,2023, 44(7):250-254.

[34] 张炳星,高军伟.基于机器视觉的金属工件智能分拣系统设计[J].工具技术,2023,57(3): 140-144.

[35] 陈继清,王志奎,强虎,等.基于机器视觉边缘检测的园林喷药机器人导航线提取[J].中国农机化学报,2021,42(3):40-47.

[36] 周竹,黄懿,李小昱,等.基于机器视觉的马铃薯自动分级方法[J].农业工程学报,2012, 28(7):178-183.

[37] 初广丽,张伟,王延杰,等.基于机器视觉的水果采摘机器人目标识别方法[J].中国农机化学报,2018,39(2):83-88.

[38] 宋怀波,何东健,辛湘俊.基于机器视觉的非结构化道路检测与障碍物识别方法[J].农业工程学报,2011,27(6):225-230.

[39] 余俊宇,刘孙俊,许桃.融合注意力机制的YOLOv7遥感小目标检测算法研究[J].计算机工程与应用,2023,59(20):167-175.

[40] 孙志青.基于深度学习的铝土矿分选方法研究[D].太原,中北大学,2021.

[41] 倪冬.基于视觉的矿用输送带纵向撕裂检测系统研究[D].徐州,中国矿业大学,2024.

[42] 李博,王学文,庞尚钟,等.煤与矸石图像特征分析及试验研究[J].煤炭科学技术,2022, 50(8):236-246.

[43] 于国防,邹士威,秦聪.图像灰度信息在煤矸石自动分选中的应用研究[J].工矿自动化,2012,38(2):36-39.

[44] 王家臣,李良晖,杨胜利.不同照度下煤矸图像灰度及纹理特征提取的实验研究[J].煤炭学报,2018,43(11):3051-3061.

[45] 李曼,段雍,曹现刚,等.煤矸分选机器人图像识别方法和系统[J].煤炭学报,2020, 45(10):3636-3644.

[46] SUN J P, SU B. Coal－rock interface detection on the basis of im-age texture features［J］. International Journal of Mining Science and Technology, 2013,23(5):681－687.

[47] 陈浜.基于视觉计算的煤岩识别方法研究[D].北京,中国矿业大学(北京),2018.

[48] 吴利刚,陈乐,吕媛媛,等.基于轻量化的输送带块煤实时监测方法[J/OL].煤炭科学技术:1-10[2024-04-07].

[49] 韦小龙,王方田,刘超,等.基于CSPNet-YOLOv7目标检测算法的煤矸图像识别模型[J/OL].煤炭科学技术:1-13[2024-03-31].

[50] 曹现刚,郝朋英,王鹏,等.多因素光照条件下高质量煤矸图像获取方法研究[J].煤炭科学技术,2023,51(1):455-463.

[51] 李娟莉,魏代良,李博,等.基于深度学习轻量化的改进SSD煤矸快速分选模型[J].东北大学学报(自然科学版),2023,44(10):1474-1480.

[52] 陈彪.基于深度学习的煤矸图像识别研究[D].徐州,中国矿业大学,2023.

[53] 刘博文.综放工作面煤矸智能识别关键技术研究[D].徐州,中国矿业大学,2022.

[54] 贺海涛,王佳豪,张海峰,等.基于U-Net的放煤状态控制关键技术研究[J].煤炭科学技术,2022,50(S2):237-243.

[55] 单鹏飞,孙浩强,来兴平,等.基于改进Faster R-CNN的综放煤矸混合放出状态识别方法[J].煤炭学报,2022,47(3):1382-1394.

[56] 谢和平,高峰,鞠杨,等.深地煤炭资源流态化开采理论与技术构想[J].煤炭学报,2017, 42(3):547-556.

[57] 王国法,任世华,庞义辉,等.煤炭工业“十三五”发展成效与“双碳”目标实施路径[J]. 煤炭科学技术,2021,49(9):1-8.

[58] 李相迪,黄英,张培晴,等.红外成像系统及其应用[J].激光与红外,2014,44(3):229-234.

[59] 徐广通,袁洪福,陆婉珍.现代近红外光谱技术及应用进展[J].光谱学与光谱分析,2000, 20(2):134-142.

[60] 代具亭,汤心溢,王世勇,等.扫描型红外焦平面探测器图像实时传输系统[J].激光与红外, 2016, 46(4): 476-480.

[61] 麦绿波.焦平面热像仪的发展与应用综述[J].红外技术,2006,28(9):497-502.

[62] 金伟其,刘斌,范永杰.红外图像细节增强技术研究进展[J].红外与激光工程,2011, 40(12):2521-2527.

[63] 刘梦,艾青,王帅,等.基于辐射加热源的材料高温发射率测量方法[J].工程热物理学报,2021,42(3):740-744.

[64] 朱志星,叶林,任宏宇,等.基于 RM-6A 红外热敏电阻的发射率在线测量系统研究[J].仪表技术与传感器,2020(11):74-78+90.

[65] 李云红,孙晓刚,原桂彬.红外热像仪精确测温技术[J].光学精密工程,2007 (9):1336-1341.

[66] 张志强,王萍,于旭东,等.高精度红外热成像测温技术研究[J].仪器仪表学报,2020,41 (5):10-18.

[67] 吴燕燕,罗铁苟,黄杰,等.基于红外热像仪测温原理的物体表面发射率计算[J].直升机技术,2011(4):25-29.

[68] Terada D, Iuchi T. Emissivity-compensated radiation thermometry method using a concave mirror[J]. The Review of scientific instruments, 2021,92(5).

[69] 宋健,郝小鹏,原遵东,等.基于控制环境辐射的黑体辐射源发射率测量方法研究[J].中国激光,2015,42(9):277-283.

[70] 关威,刘建梅,王琦,等.基于红外辐射计的物体光谱发射率测量方法[J].火力与指挥控制,2016,41(11):163-166+170.

[71] 段英.红外测温目标光谱发射率特性研究[D].成都,电子科技大学,2020.

[72] 杨立,杨桢.红外热成像测温原理与技术[M].北京,科学出版社,2012.

[73] 雷晓杰,顾国华,隋修宝,等.红外双边滤波时域高通非均匀性校正[J].激光与红外,2012,42(7):831-836.

[74] D.B. Zhou, D.J. Wang, L.J. Huo, et al. Scene-based nonuniformity correction for airborne point target detection systems[J]. Optics Express, 2017,25(13):14210-14226.

[75] 李佳,秦翰林,延翔,等.时空域非线性滤波红外序列图像非均匀性校正[J].光子学报,2015,44(4):87-91.

[76] 宋岩峰,邵晓鹏,徐军.基于双平台直方图的红外图像增强算法[J].红外与激光工程,2008,37(2):308-311.

[77] Liu N, Zhao D. Detail enhancement for high-dynamic-range infrared images based on guided image filter[J]. Infrared Physics & Technology, 2014,67:138-147.

[78] 李毅,张云峰,年轮,等.尺度变化的 Retinex 红外图像增强[J].液晶与显示,2016,31(1): 104-111.

[79] 陈芳芳,耿蕊,吕勇.激光红外大气传输透过率模型研究[J].红外技术,2015,37(6):496-501.

[80] C. Gonzalez, E. Woods 著,阮秋琦等译.数字图像处理（第二版）[M].北京:电子工业出版社,2007:59-321

[81] 刘恒辉.红外图像处理算法研究及其 FPGA 实现[D].武汉,武汉理工大学,2008.

[82] 刘陈瑶,胡梦竹,张龙飞.基于改进 CLAHE 的 SF_6 红外图像增强[J].光学技术,2021, 47(1):107⁃112.

[83] 孙冬梅,陆剑锋,张善卿.一种改进 CLAHE 算法在医学试纸条图像增强中的应用[J].中国生物医学工程学报,2016,35(4):502⁃506.

[84] 王红茹,李瑞,王佳.基于改进 CLAHE 的水下彩色图像增强算法[J].舰船电子工程,2019,39(11):119⁃122,141.

[85] 黄勇.基于双边滤波和改进 CLAHE 算法的低照度图像增强研究[D].湘潭,湘潭大学,2019.

[86] 祝志慧,何昱廷,李沃霖,等.基于改进 YOLOv7 模型的复杂环境下鸭蛋识别定位[J].农业工程学报,2023,39(11):274-285.

[87] WANG C, BOCHKOVSKIY A, LIAO H. YOLOv7: Trainablebag-of-freebies sets new state-of-the-art for real-time objectdetectors[C]//Proceedings of the IEEE/CVF Conference onComputer Vision and Pattern Recognition, 2023:7464-7475.

[88] 王金鹏,周佳良,张跃跃,等.基于优选 YOLOv7 模型的采摘机器人多姿态火龙果检测系统[J].农业工程学报,2023,39(8):276-283.

[89] 宋怀波,马宝玲,尚钰莹,等.基于 YOLOv7-ECA 模型的苹果幼果检测[J].农业机械学报,2023,54(6):233-242.

[90] 李志华,于杨.基于检测的多目标跟踪算法综述[J].物联网技术,2021,11(4):20-24.

[91] 贾豆豆.基于 YOLOv5+DeepSort 的小目标跟踪方法研究[D].太原,中北大学,2022.

[92] 陈紫强,张雅琼.一种基于 YOLOv4 的改进 DeepSort 目标跟踪算法[J].桂林电子科技大学学报,2021,41(2):140-145.