煤矿智能化是我国煤炭工业高质量发展的核心技术支撑和必然方向。煤矸识别作为煤矿智能化的重要技术，一直是热点研究问题。本文研究深度学习，基于YOLOv7目标检测算法对煤矸进行识别，主要工作如下：

    （1） 针对计算资源充足，煤矸识别精度要求较高的应用场景，提出了FN-YOLOv7煤矸识别算法。首先，采用K-means++算法重新聚类先验锚框，提升模型的收敛速度和精度；其次，引入FReLU激活函数弥补模型空间不敏感的缺陷；然后，加入规范化注意力机制提高模型整体性能；最后，采用EIOU函数计算边框回归损失。实验结果表明，FN-YOLOv7可以更加准确地识别不同光照条件下的煤矸目标，并有效降低了漏检、误检概率。

     （2）针对计算资源匮乏，煤矸识别精度要求较低的应用场景，提出了MD-YOLOv7-tiny煤矸识别算法。首先，采用MobileNetV3-Small轻量化网络主干，并引入深度可分离卷积对模型进一步压缩；其次，采用K-means++算法重新聚类先验锚框；最后，使用BiFPN结构进行特征融合。实验结果表明，在保证模型精度的条件下，MD-YOLOv7-tiny的模型参数量、浮点运算数和模型体积相较于YOLOv7-tiny均有大幅下降。将该模型部署至边缘计算平台，其推理速度可满足实时性要求。

         Coal mine intelligence is the core technical support and inevitable direction of the high-quality development of China's coal industry. As an important technology of coal mine intelligence, the identification of coal and gangue has always been a hot research issue. This thesis studies deep learning and uses YOLOv7 target detection algorithm to identify coal and gangue. The main work is as follows.

         (1) Aiming at the application scenarios with sufficient computing resources and high requirements for coal and gangue recognition accuracy, the FN-YOLOv7 algorithm is proposed. Firstly, K-means++ reclustered the prior anchor boxes for faster convergence and accuracy. Secondly, FReLU is introduced to compensate for the spatial insensitivity of the model. Then, normalization-based attention mechanism enhanced the model's overall performance. Finally, EIOU function calculated bounding box regression loss. Experimental results show that FN-YOLOv7 can more accurately identify coal and gangue targets under different lighting conditions, and effectively reduce the probability of missed detection and false detection.

         (2) Aiming at the application scenarios with lack of computing resources and low requirements for coal and gangue recognition accuracy, the MD-YOLOv7-tiny algorithm is proposed. Firstly, MobileNetV3-Small lightweight network backbone was used, and depthwise separable convolution was introduced to further compress the model. Secondly, K-means++ reclustered the prior anchor boxes. Finally, BiFPN was used for feature fusion. Experimental results show that compared with YOLOv7-tiny, MD-YOLOv7-tiny significantly reduces the params, GFLOPs, and model volume of the model. When the model is deployed to the edge computing platform, its inference speed fulfills real-time needs.

[1] 袁亮,吴劲松,杨科.煤炭安全智能精准开采关键技术与应用[J].采矿与安全工程学报,2023,40(05):861-868.

[2] 王国法.煤矿智能化最新技术进展与问题探讨[J].煤炭科学技术,2022,50(01):1-27.

[3] Zhao Y, Wang S, Cheng G, et al. Study on coal and gangue recognition method based on the combination of X-ray transmission and diffraction principle[J]. Energy Sources, Part A: Recovery, Utilization, and Environmental Effects, 2022, 44(4): 9716-9728.

[4] 葛世荣,胡而已,裴文良. 煤矿机器人体系及关键技术[J]. 煤炭学报,2020,45(1):455-463.

[5] He Lei, et al. Shape selection recognition and scattering distribution prediction of adhesion targets in multi-scale dual-energy X-ray images of coal and gangue[J]. International Journal of Coal Preparation and Utilization, 2023, 43(9): 1561-1582.

[6] 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(08):236-246.

[7] Jiang Junhao, et al. Recognition and sorting of coal and gangue based on image process and multilayer perceptron[J]. International Journal of Coal Preparation and Utilization 2023,43 (1): 54-72.

[8] Guo Y, Wang, X, et al. Research on coal and gangue recognition method based on TW-RN optimized CNN[J]. Coal Science and Technology 2022, 50(1): 228-236.

[9] 邱锡鹏. 神经网络与深度学习[M]. 北京: 机械工业出版社, 2020: 4-7.

[10] Suykens J, Vandewalle J. Least Squares Support Vector Machine Classifiers[J]. Neural Processing Letters, 1999, 9(3):293-300.

[11] Hinton, Geoffrey E, Ruslan R. Salakhutdinov. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786): 504-507.

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

[13] 何敏,王培培,蒋慧慧.基于SVM和纹理的煤和煤矸石自动识别[J].计算机工程与设计,2012,33(03):1117-1121.

[14] 袁小翠,吴禄慎,陈华伟.基于Otsu方法的钢轨图像分割[J].光学精密工程, 2016, 24(07): 1772-1781.

[15] 吴开兴,宋剑.基于灰度共生矩阵的煤与矸石自动识别研究[J].煤炭工程, 2016,48(02):98-101.

[16] Kai L, Xi Z, Chen Y Q. Extraction of Coal and Gangue Geometric Features with Multifractal Detrending Fluctuation Analysis[J]. Applied Sciences, 2018, 8(3): 463.

[17] 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.

[18] 庞尚钟,李博,王学文等.基于机器视觉的煤矸识别系统设计及试验研究[J].煤炭工程,2021,53(02):141-146.

[19] 田冬艳,丁苏凡,郭星歌.基于图像处理的煤矸识别方法[J].煤炭技术,2022,41(03):201-204.

[20] 李瑞,李博,王学文等.基于XGBoost与可见-近红外光谱的煤矸识别方法[J].光谱学与光谱分析,2022,42(09):2947-2955.

[21] 程刚,陈杰,何磊.基于LBP特征与SVM的煤矸识别方法研究[J].煤炭技术,2023,42(10):12-15.

[22] 徐志强,吕子奇,王卫东等.煤矸智能分选的机器视觉识别方法与优化[J].煤炭学报,2020,45(06):2207-2216.

[23] 郜亚松,张步勤,郎利影.基于深度学习的煤矸石识别技术与实现[J].煤炭科学技术,2021,49(12):202-208.

[24] 韩存地,朱兴攀,符立梅等.改进空间通道注意力与残差融合的煤矸石识别[J].西安科技大学学报,2021,41(06):1113-1121.

[25] 郭永存,张勇,李飞等.嵌入空洞卷积和批归一化模块的智能煤矸识别算法[J].矿业安全与环保,2022,49(03):45-50.

[26] 王闰泽,郎利影,席思星.用于智能煤矸分选机器人的改进型VGG网络煤矸识别模型[J].煤炭技术,2022,41(01):237-241.

[27] 杜京义,史志芒,郝乐等.轻量化煤矸目标检测方法研究[J].工矿自动化, 2021,47(11):119-125.

[28] Zhang Y, Wang J, Yu Z, et al. Research on intelligent detection of coal gangue based on deep learning[J]. Measurement, 2022, 198: 111415.

[29] Li D, Wang G, Wang S, et al. Research on Coal Gangue Detection and Recognition Based on Lightweight Network MS-YOLOV3[J]. Gospodarka Surowcami Minerally, 2022, 38(4): 133-152.

[30] Yan P, Kan X, Zhang H, et al. Target Recognition of Coal and Gangue Based on Improved YOLOv5s and Spectral Technology[J]. Sensors, 2023, 23(10): 4911.

[31] 沈科,季亮,张袁浩等.基于改进YOLOv5s模型的煤矸目标检测[J].工矿自动化,2021,47(11):107-111+118.

[32] 张磊,王浩盛,雷伟强等.基于YOLOv5s-SDE的带式输送机煤矸目标检测[J].工矿自动化,2023,49(04):106-112.

[33] 周志华. 机器学习[M]. 北京: 清华大学出版社, 2016: 97-99.

[34] Fan E. Extended tanh-function method and its applications to nonlinear equations[J]. Physics Letters A, 2000, 277(4-5): 212-218.

[35] MAAS A L, HANNUN A Y, NG A Y. Rectifier nonlinea-rities improve neural network acoustic models[C]//Proceedings of the 30th International Conference on Machine Learning Atlanta. ACM, 2013: 456-462.

[36] 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.

[37] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J].IEEE Transactions on Geoscience & Remote Sensing, 2015, 12: 64-73.

[38] He K, Zhang X, Ren S, et al. Deep Residual Learning for Image Recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778.

[39] Jia D, Wei D, Socher R, et al. ImageNet: A large-scale hierarchical image database[J]. Proc of IEEE Computer Vision & Pattern Recognition, 2009:248-255.

[40] Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2014: 580-587.

[41] 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.

[42] 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.

[43] Redmon J, Farhadi A. YOLO9000: better, faster, stronger[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 7263-7271.

[44] Howard A G, Zhu M, Chen B, et al. Mobilenets: Efficient convolutional neural networks for mobile vision applications[DB/OL]. https://arxiv.org/abs/1704.04861, 2017-04-17.

[45] Zhang X, Zhou X, Lin M, et al. Shufflenet: An extremely efficient convolutional neural network for mobile devices[C]. Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 6848-6856.

[46] Han K, Wang Y, Tian Q, et al. Ghostnet: More features from cheap operations[C]// Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 1580-1589.

[47] Wang C Y, Bochkovskiy A, Liao H Y M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023: 7464-7475.

[48] 赵春华,罗顺,谭金铃等.基于PC-YOLOv7算法钢材表面缺陷检测[J].国外电子测量技术,2023,42(09):137-145.

[49] 倪昌双,李林,罗文婷等.改进YOLOv7的沥青路面病害检测[J].计算机工程与应用,2023,59(13):305-316.

[50] 胡欣,周运强,肖剑,等.基于改进YOLOv5的螺纹钢表面缺陷检测[J].图学学报,2023,44(03):427-437.

[51] 黄海生, 饶雪峰. 面向无人机航拍场景的轻量化目标检测[J].计算机系统应用, 2022, 31(12):159-168.

[52] MA N N, ZHANG X Y, SUN J. Funnel activation for visual recognition[C]//Proceedings of the European Conference on Computer Vision. 2020: 351-368.

[53] Liu Y, Shao Z, Teng Y, et al. NAM: Normalization-based attention module [DB/OL]. http://arxiv.org/abs/2111.12419, 2021-12-24.

[54] 戴楚舒,张选德,熊静.基于规范化注意力机制的孪生单目标视觉追踪[J].陕西科技大学学报,2023,41(01):166-173.

[55] Zhang Y F, Ren W, Zhang Z, et al. Focal and efficient IOU loss for accurate bounding box regression [J]. Neurocomputing, 2022, 506:146-157.

[56] Tan M, Pang R, Le Q V. Efficientdet: Scalable and efficient object detection[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 10781-10790.