煤炭异物的高效检测是智能化干法选煤的关键环节，也是提升原煤入选率的重要方法。目前，国内外学者在煤炭异物检测方面成果丰硕，所讨论的问题多集中在实时成像质量低、多目标检测不准确，以及局部遮挡检测效果差等方面。本文提出了一种多目标煤炭异物分割模型，以期实现煤炭异物的高效检测与智能分割。

针对在线选煤过程中，视觉检测系统因光照条件不足引起异物表面特征成像质量差、识别率低的问题，研究融合光源分布、色温、光照强度等多因素条件的异物高质量图像获取方法。该方法通过对不同因素进行参数调节，并通过多组实验进行效果验证，最终确定出最优光源，优化后的光源大幅度提高了煤炭异物识别率。

针对煤炭异物材质、形状的多样性引起同类目标边界框区分度下降、识别准确率下降的问题，研究基于实例分割的煤炭异物检测方法。该方法引入掩膜提升目标位置与轮廓的表达，通过超参数优化、迁移学习等方式进行模型优化，最终构建煤炭异物检测模型并实现特征图与检测结果可视化，优化后的异物检测模型有效提高了多目标检测的准确率。

针对煤炭异物局部遮挡引起部分特征缺失、识别率降低的问题，研究基于改进Mask R-CNN的煤炭异物检测方法。该方法在特征获取阶段引入注意力机制CBAM强化遮挡物体可见部分的特征，在边界框回归阶段以k-means聚类算法优化锚点框比例，并且通过CIoU损失函数提升异物定位精度。实验结果表明，改进算法在目标遮挡场景下具有较好的检测性能。

为了提高煤炭异物识别率，本文提出了光源优化方法，并针对目标遮挡问题提出了基于特征获取与边界框回归优化的异物识别模型。实验结果表明：优化后的检测系统能够提高图像质量；提出的基于实例分割的煤炭异物识别方法，能够有效解决多目标识别和目标遮挡问题，并且在一定程度上提高了煤炭异物识别率。

Efficient detection of coal foreign matter is a key link in intelligent dry coal preparation and an important method to improve the selection rate of raw coal. At present, scholars at home and abroad have achieved fruitful results in coal foreign body detection, and the problems discussed are mostly focused on low quality of real-time imaging, inaccurate detection of multiple targets, and poor detection of local occlusion, etc. In this paper, a multi-objective coal foreign body segmentation model is proposed to achieve efficient detection and intelligent segmentation of coal foreign bodies.

Aiming at the problem of poor imaging quality and low recognition rate of foreign body surface features caused by insufficient lighting conditions in the visual inspection system in the process of online coal preparation, a high-quality image acquisition method of foreign bodies that integrates multiple factors such as light source distribution, color temperature, and light intensity is studied. The method finally determines the optimal light source by adjusting the parameters of different factors and verifying the effect through several sets of experiments, and the optimized light source significantly improves the recognition rate of foreign objects.

Aiming at the problem that the diversity of materials and shapes of coal foreign objects causes the decrease in the differentiation of similar target bounding boxes and the decrease in recognition accuracy, the coal foreign object detection method based on instance segmentation is studied. The method introduces a mask to enhance the expression of target location and contour, optimizes the model by hyperparameter optimization and migration learning, and finally constructs a coal foreign object detection model and visualizes the feature map and detection results, and the optimized foreign object detection model effectively improves the accuracy of multi-target detection.

Aiming at the problem of partial feature loss and recognition rate reduction caused by local occlusion of coal foreign objects, a coal foreign object detection method based on improved Mask R-CNN is investigated. The method introduces the attention mechanism CBAM to strengthen the features of the visible part of the occluded object in the feature acquisition stage, optimizes the anchor box ratio with k-means clustering algorithm in the bounding box regression stage, and improves the foreign object localization accuracy by CIoU loss function. The experimental results show that the improved algorithm has better detection performance in the target occlusion scene.

In order to improve the recognition rate of coal foreign objects, this paper proposes a light source optimization method, and proposes a foreign object recognition model based on feature acquisition and bounding box regression optimization for the target occlusion problem. The experimental results show that the optimized detection system can improve the image quality; the proposed coal foreign body recognition method based on instance segmentation can effectively solve the problems of multi-target recognition and target occlusion, and to a certain extent, improve the coal foreign body recognition rate.

[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(07): 1949-1960.

[2] 刘峰, 郭林峰, 赵路正. 双碳背景下煤炭安全区间与绿色低碳技术路径[J]. 煤炭学报, 2022, 47(01): 1-15.

[3] 卫小芳, 王建国, 丁云杰. 煤炭清洁高效转化技术进展及发展趋势[J]. 中国科学院院刊, 2019, 34(04): 409-416.

[4] 王国法, 赵国瑞, 任怀伟. 智慧煤矿与智能化开采关键核心技术分析[J]. 煤炭学报, 2019, 44(01): 34-41.

[5] 杨晨光, 冯岸岸, 朱金波, 等. 智能分选中煤矸X射线识别技术的研究[J]. 安徽化工, 2020, 46(03): 25-29+33.

[6] 王锐, 桂志国, 刘祎, 等. 基于X射线和结构光相机的煤矸石分拣方法研究[J]. 中北大学学报(自然科学版), 2021, 42(02): 123-128+134.

[7] 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.

[8] 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.

[9] 李曼, 杨茂林, 刘长岳, 等. 基于图像的煤矸分选中图像照度调节方法研究[J/OL]. 煤炭学报: 1-8[2022-03-16].

[10]Li M, He X, Duan Y, et al. Experimental study on the influence of external factors on image features of coal and gangue[J]. International Journal of Coal Preparation and Utilization, 2021: 1-18.

[11]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.

[12]商德勇, 章林, 牛艳奇, 等. 煤矸分拣机器人设计与关键技术分析[J/OL]. 煤炭科学技术:1-7[2022-04-24].

[13]曹现刚, 吴旭东, 王鹏, 等. 面向煤矸分拣机器人的多机械臂协同策略[J]. 煤炭学报, 2019, 44(S2): 763-774.

[14]王鹏, 曹现刚, 夏晶, 等. 基于机器视觉的多机械臂煤矸石分拣机器人系统研究[J]. 工矿自动化, 2019, 45(09): 47-53.

[15]曹现刚, 李莹, 王鹏, 等. 煤矸石识别方法研究现状与展望[J]. 工矿自动化, 2020, 46(01): 38-43.

[16]曹现刚, 刘思颖, 王鹏, 等. 面向煤矸分拣机器人的煤矸识别定位系统研究[J]. 煤炭科学技术, 2022, 50(01): 237-246.

[17]刘京领. 在运煤皮带机上清除煤中非铁器杂物的新方法[J]. 能源技术与管理, 2005(02): 41-87.

[18]Zheng G F, Zha J W, Liu H Z, et al. Research on a Coal Gangue Identification Device[C]. Applied Mechanics & Materials, 2013, 325-326:699-702.

[19]师永春. 煤炭杂物情况及除杂工作分析[J]. 科技情报开发与经济, 2003(10): 166-167.

[20]张俊峰. 胶带输送机上清除煤中杂物的方法[J]. 中州煤炭, 2005(05): 47-48.

[21]纪贵军. 火力发电厂煤中杂物分离机构的设计方法研究[D]. 杭州: 浙江大学, 2003.

[22]何永彬, 母海龙, 田娜, 等. 补连塔选煤厂杂物管控的措施[J]. 陕西煤炭, 2019, 38(S1): 146-148.

[23]杨春宇. 关于吕家坨矿洗煤厂块精煤机选除杂工艺的研究[J]. 中国资源综合利用, 2019, 37(09): 184-186.

[24]张军. 选煤厂杂物管控探讨[J]. 内蒙古煤炭经济, 2019(12): 87-88+94.

[25]吕志强. 复杂环境下煤矿皮带运输异物图像识别研究[D]. 徐州: 中国矿业大学, 2020.

[26]吴守鹏. 基于机器视觉的运煤皮带异物识别方法研究[D]. 徐州: 中国矿业大学, 2019.

[27]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.

[28]王卫东, 张康辉, 吕子奇, 等. 基于机器视觉的煤中杂物智能分选系统研究[J]. 选煤技术, 2020(02): 87-91.

[29]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+90.

[30]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/OL]. 控制工程: 1-8[2022-03-16].

[31]王卫东, 张康辉, 吕子奇, 等. 基于深度学习的煤中异物机器视觉检测[J]. 矿业科学学报, 2021, 6(01): 115-123.

[32]尚会超, 杨锐, 段梦珍, 等. 机器视觉照明系统的关键技术分析[J]. 中原工学院学报, 2016, 27(03): 16-21.

[33]王晓旭. 模压镜片疵病检测照明系统研究[D]. 长春: 长春理工大学, 2020.

[34]孔祥伟. 组合光源与图像处理算法在工件表面缺陷检测中的应用[D]. 天津:天津大学, 2007.

[35]史晨阳, 林燕丹. 机器视觉中均匀照明成像的LED环形阵列设计[J]. 光学学报, 2020, 40(20): 118-128.

[36]冯波, 汤伟, 单文娟. 基于近场均匀照明的纸病检测系统光源优化方案[J]. 中国造纸, 2020, 39(03): 58-65.

[37]周广增, 滕奇志, 吴小平, 等. 透射光源下提升图像对比度的照明系统[J]. 光电子.激光, 2012, 23(03): 595-601.

[38]杨锐. 机器视觉照明技术与装置实验研究[D]. 郑州: 中原工学院,2016.

[39]蒙春学. 理管机机器视觉光源关键技术研究[D]. 西安: 西安工程大学, 2019.

[40]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(02): 36-39.

[41]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.

[42]龚波. 基于深度学习的目标检测和实例分割算法研究[D]. 成都: 西南民族大学, 2021.

[43]王海波. 基于深度学习的单阶段实例分割方法研究[D]. 徐州:中国矿业大学, 2021.

[44]苏丽, 孙雨鑫, 苑守正. 基于深度学习的实例分割研究综述[J]. 智能系统学报, 2022, 17(01): 16-31.

[45]张继凯, 赵君, 张然, 等. 深度学习的图像实例分割方法综述[J]. 小型微型计算机系统, 2021, 42(01): 161-171.

[46]Ren S, He K, Girshick R, et al. Faster r-cnn: Towards real-time object detection with region proposal networks[J]. Advances in neural information processing systems, 2015, 28.

[47]He K, Gkioxari G, Dollár P, et al. Mask r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2961-2969.

[48]Chen L C, Hermans A, Papandreou G, et al. Masklab: Instance segmentation by refining object detection with semantic and direction features[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 4013-4022.

[49]Liu S, Qi L, Qin H, et al. Path aggregation network for instance segmentation[C]// Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 8759-8768.

[50]Huang Z, Huang L, Gong Y, et al. Mask scoring r-cnn[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019: 6409-6418.

[51]Bolya D, Zhou C, Xiao F, et al. Yolact: Real-time instance segmentation[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2019: 9157-9166.

[52]Bolya D, Zhou C, Xiao F, et al. Yolact++: Better real-time in-stance segmentation[J]. arXiv preprint arXiv: 1912.06218, 2019．

[53]Chen H, Sun K, Tian Z, et al. Blendmask: Top-down meets bottom-up for instance segmentation[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 8573-8581.

[54]Liu S, Jia J, Fidler S, et al. Sgn: Sequential grouping networks for instance segmentation [C]//Proceedings of the IEEE International Conference on Computer Vision. 2017: 3496-3504.

[55]Chopra S, Hadsell R, LeCun Y. Learning a similarity metric discriminatively, with application to face verification[C]//2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05). IEEE, 2005, 1: 539-546.

[56]Gao N, Shan Y, Wang Y, et al. Ssap: Single-shot instance segmentation with affinity pyramid[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019: 642-651.

[57]Wang X, Kong T, Shen C, et al. Solo: Segmenting objects by locations[C]//European Conference on Computer Vision. Springer, Cham, 2020: 649-665.

[58]Wang X, Zhang R, Kong T, et al. Solov2: Dynamic, faster and stronger[J]. arXiv e-prints, 2020: arXiv: 2003. 10152.

[59]王绍清, 常方哲, 陈昊, 等. 高变质煤HRTEM图像中芳香晶格条纹的MASK R-CNN识别[J]. 煤炭学报, 2021, 46(02): 591-601.

[60]曹现刚, 费佳浩, 王鹏, 等. 基于多机械臂协同的煤矸分拣方法研究[J]. 煤炭科学技术, 2019, 47(04): 7-12.

[61]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.

[62]逯圣辉. 基于机器视觉的带式输送机动态煤量计量研究[D]. 邯郸:河北工程大学, 2020.

[63]曹现刚, 郝朋英, 王鹏, 等. 多因素光照条件下高质量图像获取方法研究[J/OL]. 煤炭科学技术: 1-11[2022-02-11].

[64]浦昭邦, 屈玉福, 王亚爱. 视觉检测系统中照明光源的研究[J]. 仪器仪表学报, 2003(S2): 438-439.

[65]于丽杰, 李德胜, 董南萍. 视觉检测中的光照强度优化研究[J]. 计算机工程与应用, 2012, 48(17): 178-182+192.

[66]芮大为. LED微型投影光源的二次光学设计[D]. 成都: 电子科技大学, 2012.

[67]李培强, 薛文琦, 唐学贤, 等. 一种基于TOPSIS法的光伏用户群内用户重要性评估方法[J]. 湖南大学学报(自然科学版), 2020, 47(12): 117-123.

[68]傅泰铭. 基于卷积神经网络的实例分割算法研究[D]. 南宁: 广西大学, 2021.

[69]Lin T Y, Maire M, Belongie S, et al. Microsoft coco: Common objects in context[C] //European conference on computer vision. Springer, Cham, 2014: 740-755.

[70]Cordts M, Omran M, Ramos S, et al. The cityscapes dataset for semantic urban scene understanding[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 3213-3223.

[71]Neuhold G, Ollmann T, Rota Bulo S, et al. The mapillary vistas dataset for semantic understanding of street scenes[C]//Proceedings of the IEEE international conference on computer vision. 2017: 4990-4999.

[72]Gupta A, Dollar P, Girshick R. LVIS: A dataset for large vocabulary instance segmentation[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 5356-5364.

[73]高烁琪. 齿轮箱磨粒铁谱图像智能分类与异常检测研究[D]. 西安: 西安科技大学, 2021.

[74]肖力炀. 基于改进型Mask R-CNN模型的路面裂缝自动化识别方法[D]. 西安:长安大学, 2021.

[75]张胜虎, 马惠敏. 遮挡对于目标检测的影响分析[J]. 图学学报, 2020, 41(06): 891-896.

[76]杜先君, 巩彬, 余萍, 等. 基于CBAM-CNN的模拟电路故障诊断研究[J/OL]. 控制与决策: 1-9[2022-03-17].

[77]Jaderberg M, Simonyan K, Zisserman A. Spatial transformer networks[J]. Advances in neural information processing systems, 2015, 28.

[78]Gao P, Zheng M, Wang X, et al. Fast convergence of detr with spatially modulated co-attention[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2021: 3621-3630.

[79]王瑶. 基于改进的Faster R-CNN的快递识别系统研究[D]. 南京: 南京邮电大学, 2021.

[80]Yu J, Jiang Y, Wang Z, et al. Unitbox: An advanced object detection network [C]//Proceedings of the 24th ACM international conference on Multimedia. 2016: 516-520.

[81]Rezatofighi H, Tsoi N, Gwak J Y, et al. Generalized intersection over union: A metric and a loss for bounding box regression[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 658-666.

[82]Zheng Z, Wang P, Liu W, et al. Distance-IoU loss: Faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2020, 34(07): 12993-13000.

[83]李莹. 基于深度学习的煤矸石目标检测方法研究[D]. 西安: 西安科技大学, 2020.