摘要
基于井下煤岩分界面的图像信息对工作面的煤岩分界情况进行高效识别,是实现采煤机连续自主截割和煤矿智能化无人开采的关键技术。基于图像处理技术的煤岩分界识别方法具有灵活性高和适应性强等特点,是解决煤岩分界识别问题有效且优势明显的途径。基于河北省矿山智能化开采技术重点实验室煤岩分界物理试验平台,考虑工作面正常开采情况和3种在开采过程中极易产生的干扰因素,模拟了实际开采状态下真实的煤岩分界面情况,通过处理试验平台采集的图像资源,使用3种传统机器学习的方法对煤岩分界图像进行检测识别,得到工作面煤岩分界线图像。针对传统检测方法在复杂情况下缺少足够适应性的问题,利用基于深度学习的网络模型对智能化工作面煤岩分界图像进行识别分析,形成煤与岩的预测图,得出改进后的DeeplabV3+网络模型在面对复杂情况时具有较强的适用性和稳定性。基于研究成果得出深度学习方法相较于传统机器学习方法能更好对煤岩分界面图像进行处理,识别的准确率较高,更适用于井下复杂环境的情况,是未来煤岩分界识别技术发展的重要方向。同时基于煤岩分界线识别预测结果,结合智能化开采工作面实际情况,设计了智能截割方案。
Efficiently identifying the coal-rock interface in working face based on underground image information is a key technology for continuous autonomous cutting of shearers and intelligent unmanned mining in coal mines.The coal-rock interface recognition method based on image processing technology offers high flexibility and strong adaptability,providing an effective and highly advantageous approach for solving the coal-rock interface recognition issue.Based on the physical testing platform for coal-rock interface recognition at Hebei Provincial Key Laboratory of Intelligent Mining Technology,considering the normal mining situation of the working face and three interference factors easily generated during the mining process,the real coal-rock interface situation under actual mining conditions was simulated.By processing the image resources collected by the experimental platform,the coal-rock interface images were detected and recognized by three traditional machine learning methods,obtaining the coal-rock interface images of the working face.In response to the insufficient adaptability of traditional detection methods in complex situations,a deep learning-based network model was utilized to identify and analyze the coal-rock interface images in the intelligent working face,generating prediction images of coal and rock.It was concluded that the improved DeeplabV3+network model demonstrated strong applicability and stability when handling complex conditions.Based on the research findings,it was showed that compared to traditional machine learning methods,deep learning methods were more effective in processing coal-rock interface images with higher identification accuracy,and were better suited for complex underground environments and future development direction of coal-rock interface recognition technology.On the basis of the recognition and prediction results of the coal-rock interface line and in consideration of the actual conditions of the intelligent mining face,an intelligent cutting scheme was designed.
作者
张科学
李举然
李明忠
何满潮
王晓玲
张强
孟庆勇
任怀伟
庞义辉
张立亚
刘清
冯银辉
李伟涛
ZHANG Kexue;LI Juran;LI Mingzhong;HE Manchao;WANG Xiaoling;ZHANG Qiang;MENG Qingyong;REN Huaiwei;PANG Yihui;ZHANG Liya;LIU Qing;FENG Yinhui;LI Weitao(Hebei Provincial Key Laboratory of Intelligent Mining Technology,North China University of Science and Technology,Langfang,Hebei 065201,China;Key Laboratory of Coal Mine Intelligent Safety Technology and Equipment,Ministry of Emergency Management,North China University of Science and Technology,Langfang,Hebei 065201,China;State Key Laboratory of Disaster Prevention and Control of Tunnel Engineering and Intelligent Construction and Maintenance,China University of Mining and Technology-Beijing,Haidian,Beijing 100083,China;Joint Research Academy(Smart Mine Research Academy)(Institute of Intelligent Unmanned Mining),North China Institute of Science and Technology,Langfang,Hebei 065201,China;School of Mine Safety,North China Institute of Science and Technology,Langfang,Hebei 065201,China;China Coal Technology and Engineering Group Co.,Ltd.,Chaoyang,Beijing 100013,China;College of Mechanical and Electronic Engineering,Shandong University of Science and Technology,Qingdao,Shandong 266590,China)
出处
《中国煤炭》
北大核心
2025年第12期148-162,共15页
China Coal
基金
深部岩土力学与地下工程国家重点实验室(北京)开放基金资助项目(SKLGDUEK1822)
廊坊市科技支撑计划项目(2025013166)
河北省矿山智能化开采技术重点实验室开放基金项目(iium009)
中央高校基本科研业务费资助项目(3142021007,3142019009)
中国科协科技智库青年人才计划(20220615ZZ07110397)。
关键词
智能化工作面
煤岩分界识别
图像处理
机器学习
深度学习
intelligent coal mining face
coal-rock interface recognition
image processing
machine learning
deep learning
