煤炭作为我国的能源支柱，生产安全问题不可忽视。但在煤炭生产过程中经常会混入异物，严重影响生产安全和产品质量。如何将煤中混入的异物及时识别出来成为研究热点，具有较大实用价值。本文针对传统算法识别目标单一，机器学习复杂度高等问题，提出两种适用不同场景的煤中异物识别算法。

（1）针对背景图像固定、学习数据较少的固定场景，提出一种煤中多异物同时识别的SW-SVM算法。该算法由特征提取与选择、SVM分类器优化、SW-SVM识别三个过程构成。首先提取异物与背景图像在空域和频域中的多个特征，经过相关系数法的自适应处理，选择相关度较高的部分特征。在SVM分类器优化阶段，通过对比实验确定了效果较好的SVM分类器及Ploy核函数，并用网格搜索算法对Ploy核函数的可调参数进行优化。在SW-SVM识别阶段，通过SW（滑窗）方式，用优化后的SVM分类器对滑动窗口中的子图像逐一进行检测、分配标签并进行灰度重置。最后利用形态学处理、连通区域检测和最小外接矩形完成最终的异物标注。这种利用滑窗思想的SW-SVM算法有效提高了异物识别的精度。该算法不仅实现简单，还可以实现多种异物的同时识别。对输煤皮带上出现三种不同异物的情景进行了仿真实验，效果较好。

（2）针对背景图像不定、学习数据较多的非固定场景，提出一种YOLOX的改进算法RepVGG-YOLOX。首先使用RepVGG Block、SPP、深度可分离卷积等重新设计YOLOX的特征提取层和融合层网络结构。然后对YOLOX的解耦检测头进行轻量化，减少模型参数量，同时加快推理速度。最后将YOLOX的损失函数IOU替换为CIOU。实验结果表明，RepVGG-YOLOX相对于YOLOX模型，mAP从97.44%提升到98.0%，FPS提升了61%，模型参数量和浮点运算数分别减少了23%和25%。本文改进的RepVGG-YOLOX模型结构简洁，参数量少，实时性高，具有较大的实用价值。

As Chinese energy pillar, coal production safety issues cannot be ignored. However, in the process of coal production, foreign bodies are often mixed, which seriously affects production safety and product quality. How to identify the foreign body mixed in coal has become a research hotspot and has great practical value. Aiming at the problems of single target recognition and high complexity of machine learning in traditional algorithms, this paper proposes two foreign object recognition algorithms in coal suitable for different scenarios.

(1) For fixed scenes with fixed background images and less learning data，a recognition algorithm of multiple foreign bodies in coal based on SW-SVM is proposed. The method consists of three procedures: feature extraction and selection, SVM classifier optimization, and SW-SVM recognition. Firstly, multiple features of the foreign bodies and the background image in the spatial and frequency domains are extracted, and selects some features with high correlation coefficient through adaptive processing method. In the optimization stage of SVM classifier, the SVM classifier and Ploy kernel function with better effect are determined by comparative experiments, and the hyperparameters of Ploy kernel function are optimized by grid search algorithm. In the SW-SVM recognition stage, through the SW (sliding window), the optimized SVM classifier is used to detect the sub-images in the sliding window one by one, assign labels and reset the grayscale. Finally, morphological processing, connected region detection and minimum circumscribed rectangle are used to complete the final foreign bodies labeling. This SW-SVM algorithms using sliding window idea effectively improves the accuracy of foreign bodies recognition. This algorithm is not only simple to implement, but also can realize the simultaneous recognition of multiple foreign bodies. The simulation experiment of three different foreign bodies on the coal conveyor belt is carried out, and the effect is good.

(2) Aiming at the non-fixed scenes with uncertain background images and more learning data, an improved algorithm RepVGG-YOLOX based on YOLOX is proposed. Firstly, the feature extraction layer and fusion layer network structure of YOLOX are redesigned by using RepVGG Block, SPP and Deep Separable Convolution. Then, the Decoupled Head of YOLOX is lightweighted to reduce the number of model parameters and accelerate the reasoning speed. Finally, replaces the loss function IOU of YOLOX with CIOU. The experimental results show that compared with the YOLOX model, mAP of RepVGG-YOLOX increases from 97.44 % to 98.0 %, FPS increases by 61 %, and the number of model parameters and floating-point operations decreases by 23 % and 25 %. The improved RepVGG-YOLOX model in this paper has simple structure, less parameters, high real-time performance and great practical value.

[1] 曹虎奇. 煤矿带式输送机撕带断带研究分析[J]. 煤炭科学技术, 2015, 43(S2): 130-134.

[2] 郭永存,何磊,刘普壮,等. 煤矸双能X射线图像多维度分析识别方法[J]. 煤炭学报, 2021, 46(01): 300-309.

[3] 宋亮, 刘善军, 虞茉莉, 等. 基于可见-近红外和热红外光谱联合分析的煤和矸石分类方法研究[J]. 光谱学与光谱分析, 2017, 37(02): 416-422.

[4] 马宪民.煤矸石在线识别与自动分选系统的研究[J]. 西安科技学院学报, 2003(01):66-68.

[5] 陈立, 杜文华, 曾志强,等. 基于小波变换的煤矸石自动分选方法[J]. 工矿自动化, 2018, 44(12): 60-64.

[6] 王燕, 郭潇樯, 刘新华. 带式输送机大块异物视觉检测系统设计[J]. 机械科学与技术, 2021, 40(12): 1939-1943.

[7] 李沙,李春娟.机场跑道异物检测系统设计与算法研究[J]. 现代雷达, 2021, 43(06): 80-85.

[8] 薛旭升, 杨星云, 齐广浩, 等. 煤矿带式输送机分拣机器人异物识别与定位系统设计[J]. 工矿自动化, 2022, 48(12): 33-41.

[9] 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.

[10] 肖明尧, 李雄飞, 张小利. 基于多尺度的区域生长的图像分割算法[J]. 吉林大学学报(工学版), 2017, 47(05): 1591-1597.

[11] Feiyan B, Minqiang F, Hongli Y. Image segmentation method for coal particle size distribution analysis[J]. Particuology, 2021(3): 163-170.

[12] Faris H, Hassonah M A, Alam AZ, et al. A multi-verse optimizer approach for feature selection and optimizing SVM parameters based on a robust system architecture[J]. Neural Computing and Applications, 2018, 30(8); 2355-2369.

[13] Shichao Z, Xuelong L, Ming Z, et al. Efficient KNN Classification With Different Numbers of Nearest Neighbors[J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 29(5): 1774-1785.

[14] Subasi A, Dammas D H, Algamdl R D, et al. Sensorbased human activity recognition using adaboost ensemble classifier [J]. ProcediaComputer Science, 2018, 140: 104-111.

[15] 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.

[16] 郭永存, 于中山, 卢熠昌．基于PSO优化NP-FSVM的煤矸光电智能分选技术研究[J].煤炭科学技术, 2019, 47(04): 13-19.

[17] 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.

[18] 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.

[19] 薛光辉, 李秀莹, 钱孝玲, 等. 基于随机森林的综放工作面煤矸图像识别[J]. 工矿自动化, 2020, 46(05): 57-62.

[20] 杨奇让, 胡振琪, 韩佳政, 等. 煤矿区无人机影像采动地裂缝提取方法研究[J]. 煤炭科学技术（网络首发版）, 2022, 1-13.

[21]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.

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

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

[24] Simonyan K, Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition[J]. Computer Science, 2014, 1-14.

[25] Szegedy C, Wei L, Jia Y, et al. Going Deeper with Convolutions[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015, 1-9.

[26] Ioffe S, Szegedy C. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift[J], JMLR.org, 2015, 1-11.

[27] Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the Inception Architecture for Computer Vision[J]. Proceedings of the IEEE, 2016, 2818-2826.

[28] Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning[C]. Proceedings of the 31st AAAI Conference on Artificial Intelligence, 2016:4278-4284.

[29] He K M, Zhang X Y, Ren S Q, et al. Deep Residual Learning for Image Recognition[C]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA, 2016, 770-778.

[30] Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detectionand semantic segmentation[C]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Columbus, OH, USA, 2014, 580-587.

[31] He K M, Zhang X Y, Ren S Q, et al. Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition[J]. COMPUTER VISION-ECCV 2014, PT III, 2014, 8691: 346-361.

[32] Girshick R. Fast R-CNN[C]. Proceedings of IEEE International Conference on Computer Vision, Santiago, Chile, 2015, 1440-1448.

[33] Ren S Q, He K M, Girshick R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks[J]. Proceedings of the IEEE transactions on pattern analysis and machine intelligence, 2017, 39(6): 1137-1149.

[34] Redmon J, Divvala S, Girshick R, et al. You Only Look Once: Unified, Real-Time Object Detection[C]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 2016: 779-788.

[35] Liu W, Anguelov D, Erhan D, et al. SSW: Single Shot MultiBox Detector[J]. Lecture Notes in Computer Science, 2016, 9905: 21-37.

[36]吴守鹏，丁恩杰，俞啸. 基于改进FPN的输送带异物识别方法[J]. 煤矿安全，2019，50(12): 127-130.

[37] Yuanbin W, Yujing W, Langfei D. Video detection of foreign objects on the surface of belt conveyor underground coal mine based on improved SSD [J]. Journal of Ambient Intelligence and Humanized Computing, 2020, 1-10.

[38] Kanghui Z, Weidong W, Ziqi L. Computer vision detection of foreign objects in coal processing using attention CNN[J]. Engineering Applications of Artificial Intelligence, 2021, 102:104242.

[39]任志玲, 朱彦存. 改进CenterNet算法的煤矿皮带运输异物识别研究[J]. 控制工程, 2021:1-8.

[40]胡璟皓, 高妍, 张红娟, 等. 基于深度学习的带式输送机非煤异物识别方法[J]. 工矿自动化, 2021, 47(06): 57-62.

[41] Mengchao Z, Yueshuai C, Kai J, et al. Proactive measures to prevent conveyor belt Failures: Deep Learning-based faster foreign object detection[J]. Engineering Failure Analysis, 2022,141:106653.

[42] Qinghua M, Shikun L, Xin H, et al. Coal Mine Belt Conveyor Foreign Objects Recognition Method of Improved YOLOv5 Algorithm with Defogging and Deblurring[J]. Energies (Basel), 2022, 15(24):9504

[43] Xiangang C, Hengyang W, Peng W, et al. High Quality Coal Foreign Object Image Generation Method Based on StyleGAN-DSAD[J]. Sensors, 2023, 23(1): 374.

[44] 李渝, 吴涛, 杨文海, 等. 一种分布式目标滑窗检测方法[J]. 空间电子技术, 2022, 19(02): 64-68.

[45] Rosenfeld A. The max Roberts operator is a Hueckel-type edge detector [J]. IEEE Transactions on PatternAnalysis & Machine Intelligence, 1981, 3(1): 101-103.

[46] LANG Y, ZHENG D. An improved Sobel edge detection operator[C] IEEE International Conference on ComputerScience & Information Technology, 2010: 67-71.

[47] YANG L, WU X Y, ZHAO D W, et al. An improved Prewitt algorithm for edge detection based on noised image[C] International Congress on Image and SignalProcessing, 2011: 1197-1200.

[48] ULUPINAR F, MEDIONI G. Refining edges detected by a LoG operator [J]. Computer Vision Graphics & ImageProcessing, 1990, 51(3):275-298.

[49]LI E S, ZHU S L, ZHUBS, et al. An adaptive edge-detection method based on the Canny operator [C]. International Conference on Enoironmental Science and Information Applica- tion Technology, 2009: 465-469.

[50] ACHANTA R, SHAJI A, SMITH K, et al. SLIC superpixels compared to state-of-the-art superpixel methods[J]. IEEE Transactions on Pattern Analysis and MachineIntelligence, 2012, 34(11): 2274-2282.

[51] BOYKOV Y Y, JOLLY M P. Interactive graph cuts for optimal boundary region segmentation of objects in N-D images[C]. Proceedings of the IEEE/CVF International Conference on Computer Vision, 2001: 105-112.

[52] ROTHER C, KOLMOGOROV V, BLAKE A, et al. “GrabCut”: Interactive foreground extraction using iterated graph cuts[J]. International Conference on Computer Graphics and Interactive Techniques, 2004, 23(3): 309-314.

[53] TANG M, GORELICK L, VEKSLER O, et al. GrabCut in One Cut[C]. Proceedings of the IEEE International Conference on Computer Vision. 2013: 1769-1776.

[54] Mao J, Jain A K, Artificial neural networks for feature extraction and multivariate data projection[J]. IEEE Transactions on Neural Networks, 1995, 6(2): 296-317.

[55]薄华, 马缚龙, 焦李成. 图像纹理的灰度共生矩阵计算问题的分析[J].电子学报, 2006(01): 155-158.

[56]杨洁, 刘海民, 叶晶晶. 基于改进SIFT算子的动态障碍物检测方法[J]. 机械设计与制造, 2017(11): 270-272.

[57]张锐娟, 张建奇, 杨翠. 基于SURF的图像配准方法研究[J]. 红外与激光工程, 2009, 38(01): 160-165.

[58]李小红, 谢成明, 贾易臻, 等. 基于ORB特征的快速目标检测算法[J]. 电子测量与仪器学报, 2013, 27(05): 455-460.

[59] Chun-Li, L I, Shen L J. Texture image classification method based on improved LBP operator[J]. Computer Engineering & Design, 2016, 37(01): 232-236.

[60] Liu H, Ji H, Zhang J, et al. A novel approach for feature extraction from a gamma-ray energy spectrum based on image descriptor transferring for radionuclide identification[J]. Nuclear Science and Techniques, 2022, 33(12): 90-106.

[61]张伟, 何金国. Hu不变矩的构造与推广[J]. 计算机应用, 2010, 30(09): 2449-2452.

[62]冈萨雷斯, 伍兹, 艾丁斯. 数字图像处理的MATLAB实现(第2版)[M]. 清华大学出版社, 2013: 89-90.

[63] 王学民.概率论与数理统计[M].上海:复旦大学出版社, 2017:107.

[64] 温博文, 董文瀚, 解武杰, 等. 基于改进网格搜索算法的随机森林参数优化[J]. 计算机工程与应用, 2018, 54(10): 154-157.

[65] Ge Zheng, Liu Song-tao, Wang Feng, et al. YOLOX: Exceeding YOLO Series in 2021[J]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2021, 1-7.

[66] Lin T Y, Goyal P, Girshick R, et al. Focal Loss for Dense Object Detection[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, PP (99): 2999-3007.

[67] Tian Z, Shen C, Chen H, et al. FCOS: Fully Convolutional One-Stage Object Detection [C]. Proceedings of the IEEE/CVF International Conference on Computer Vision, 2020, 9626-9635.

[68] Yue Wu, Yinpeng Chen, Lu Yuan, et al. Rethinking classification and localization for object detection[J]. Proceedings of the IEEE/CVF Conference on omputer Vision and Patten Recognition, 2020， 10183-10192.

[69] Sandler M, Howard A, Zhu M, et al. MobileNetV2: Inverted Residuals and Linear Bottlenecks[J]. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018, 4510-4520.

[70] Ding X, Zhang X, Ma N, et al. RepVGG: Making VGG-style ConvNets Great Again[J]. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, 13728-13737.