煤矸石准确识别是确保机器人高效分拣的首要环节，但由于在实际生产中，矸石因环境污染造成的传统的图像识别率低或者无法识别，对煤矸石分拣机器人分拣效率造成严重制约。针对该问题，本文在图像识别的基础上，考虑煤矸的密度差异，提出了一种煤矸石二次识别方法，以期降低矸石的误检率。

本文分析了不同煤矿的煤与矸石同体积下的质量差，设计了基于双目视觉的煤矸二次识别方案。通过对现有不规则物体体积计算方法的研究，确定了基于双目视觉的三维重建体积计算方案，并结合煤矸分拣机器人的分拣流程及机械结构，制定了煤矸动态称重方案，以期得到准确的煤矸二次识别结果。

针对双目相机不能拍摄目标全貌导致的点云缺失问题，分析点云数据轮廓特性，利用点云翻转和补齐支撑面的方法完成了点云数据的补全，使用泊松重建法完成三维重建并计算煤矸体积，并结合矸石密度计算目标检测质量。实验证明：点云翻转方法所获得的煤矸体积更准确，误差在4%左右。

针对机械臂运动过程中因机械臂速度、加速度等引起的动态下煤矸质量波动的问题，考虑末端执行器抓取阶段对动态称重结果的影响，分析机械臂抓取过程，选择可稳态测量矸石质量阶段，建立动态称重模型，结合煤矸分拣机器人系统参数对模型进行求解。在煤矸分拣机器人平台完成动态称重实验，对比真实质量以及称重质量，并作质量测量误差分析。实验表明：称重质量最大误差为5.74%，最小误差为1.71%，且动态质量的测量时间均在0.15s内，满足工况要求。

根据煤矸体积检测方法及煤矸动态称重方法研究，本文将基于双目视觉的三维重建体积检测方法和动态质量获取方法应用于煤矸二次识别。开发煤矸二次识别程序，设计二次识别实验方案，并基于煤矸分拣机器人平台测试。本文的研究可为煤矸图像识别的准确率进一步提高提供理论基础。

Accurate identification of coal gangue is the first step to ensure the efficient sorting of robots. However, in actual production, the traditional image recognition rate of coal gangue is low or cannot be recognized due to environmental pollution, which seriously restricts the sorting efficiency of coal gangue sorting robots. Aiming at this problem, this paper puts forward a secondary identification method of coal gangue based on image recognition, considering the coal gangue density difference, in order to reduce the false detection rate of coal gangue.

In this paper, the quality difference of coal and gangue in different coal mines with the same volume is analyzed. In addition, a secondary identification scheme of coal and gangue based on binocular vision is designed. Through studying the existing irregular object volume calculation methods, a three-dimensional reconstruction scheme based on binocular vision is determined. Moreover, Combined with the sorting process and mechanical structure of the coal gangue sorting robot, a dynamic weighing scheme is formulated to obtain an accurate secondary coal gangue identification result.

Aiming at the problem that the binocular camera cannot capture the whole target picture, the point cloud data outline characteristics are analyzed. The point cloud data is completed by turning over the point cloud and filling the supporting surface. The Poisson reconstruction method is used to complete the three-dimensional reconstruction and calculate the coal gangue volume, and combined with the gangue density to calculate the target detection quality. The experiment proves that the volume of coal gangue obtained by point cloud turnover method is more accurate, and the error is about 4%.

Aiming at the problem that the dynamic coal gangue quality fluctuation caused by the manipulator speed and acceleration in its movement, considering the influence of the end effector's grabbing stage on the dynamic weighing results, the grabbing process of the manipulator is analyzed. Select the stage of steady-state measurement of gangue quality. Then, establishe a dynamic weighing model. The model is solved by the the coal gangue sorting robot system parameters. Complete the dynamic weighing experiment on the robot platform of coal gangue sorting, compare the real quality with the weighing quality, and analyze the quality measurement error. The experiment results show that the maximum error is 5.74%, while the minimum error is 1.71%. Furthermore, the measurement time of dynamic quality is within 0.15s, which meets the requirement of working condition.

According to the research of coal gangue volume detection method and coal gangue dynamic weighing method, the three-dimensional reconstruction volume detection method and dynamic quality acquisition method based on binocular vision are applyed to the coal gangue secondary identification. Develop the secondary identification program of coal gangue, design the experimental scheme of secondary identification, and test it based on the coal gangue sorting robot platform.  This research can provide a theoretical basis for improving the accuracy of coal gangue image recognition.

[1]晋城矿区煤层气井排采研究[J]. 张二娟. 能源与节能. 2017(03).

[2]MA Dan, DUAN Hongyu, LIU Jiangfeng. The role of gangue on the mitigation of mining induced hazards and environmental pollution: An experimental investigation[J]． Science of the Total Environment, 2019, 664: 436－448.

[3]中国煤炭工业改革发展年度报告( 2016年) [J]. 中国煤炭, 2017, 43( 2) : 10．Annual report on the reform and development of China’s coal indus-try( 2016) [J]．China Coal, 2017, 43( 2) : 10．

[4]石焕, 程宏志, 刘万超．我国选煤技术现状及发展趋势[J]．煤炭科学技术, 2016, 44(6):169－174．

[5]韦鲁滨, 孟丽诚, 程相锋, 李大虎, 朱学帅. 重介质悬浮液流变特性研究[J]. 煤炭学报, 2016, 41(04):992-996. DOI:10. 13225/j. cnki. jccs. 2015. 1715.

[6]BAHRAMIA, GHORBANIY, MIMOHAMMADI M, et al. The beneficiation of tailing of coal preparation plant by heavy-medium cy-clone[J]. International Journal of Coal Science & Technology, 2017, 4( 3) : 374－384．

[7]SHAHBAZI B, CHELGANIS C. Modeling of fine coal flotation separation based on particle characteristics and hydrodynamic conditions[J]. International Journal of Coal Science Technology, 2016, 3( 4): 429－439．

[8]张博, 刘爱辉．半直接浮选工艺的研究与应用效果分析[J]．选煤技术, 2017(4):19－22, 26．

[9]杨康, 娄德安, 李小乐．SKT跳汰选煤技术发展现状与展望[J]．煤炭科学技术, 2008, 36(5):1－4．

[10]ZHAO Yi ding, SUN Ming feng. Imagine processing and recognition system based on Da Vinci technology for coal and gangue[J]. Applied Mechanics and Materials, 2011, 130-134:2107-2110

[11]宋剑. 基于图像特征的煤与矸石识别算法研究[D]. 河北工程大学, 2016.

[12]周甲伟, 王福荣, 刘瑜, 杜长龙. 井下弹道式煤矸分选的理论和实验[J]. 中南大学学报(自然科学版), 2015, 46(02):498-504.

[13]岳广礼. 重介选煤工艺多信息的模糊控制方法研究[J]. 选煤技术, 2012(05):100-103. DOI:10. 16447/j. cnki. cpt. 2012. 05. 029.

[14]ZHANG Bo, LIU Aihui. Study of the semi-direct flotation process and analysis of its operating performance [J]. Coal Preparation Technology, 2017(04):19-22+26.

[15]YANG Kang, LOU Dean, LI Xiaole. Present status and outlook of SKT jig coal preparation technology[J]. Coal Science & Technology, 2008, 36(5).

[16]L Su, X Cao, H Ma, et al. Research on Coal Gangue Identification by Using Convolutional Neural Network, 2018 2nd IEEE Advanced Information Management, Communicates, Electronic and Automation Control Conference (IMCEC), Xi'an, 2018, pp. 810-814.

[17]He K, Zhang X, Ren S, et al. Deep Residual Learning for Image Recognition[C]// IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2016:770-778.

[18]费佳浩. 煤矸分栋机器人结构设计及运动分析[D]. 西安科技大学, 2019.

[19]曹现刚, 吴旭东, 王鹏, 李莹, 刘思颖, 张国祯, 夏护国. 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. DOI:10. 13225/j. cnki. jccs. 2019. 0734.

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

[21]马占国, 孙凯, 赵国贞, 潘银光, 范金泉. 煤矿井下湿法分选系统设计[J]. 煤炭科学技术, 2011, 39(02):119-121.

[22]王波, 张致维, 王然风. 重介悬浮液密度自动控制系统的设计[J]. 控制工程, 2011, 18(S1):67-69. DOI:10. 14107/j. cnki. kzgc. 2011. s1. 045.

[23]匡亚莉, 解京选, 戈军, 张军. 跳汰过程中25和13mm颗粒运动的数学模型[J]. 中国矿业大学学报, 2010, 39(06):837-842+864.

[24]高嵩, 王敏, 吴菠. TDS智能干选机在煤矿中的研究与应用[J]. 煤炭科学技术, 2020, 48(S2):57-60.

[25]郭永存, 何磊, 刘普壮, 王希. 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. DOI:10. 13225/j. cnki. jccs. 2020. 1626.

[26]VON Ketelhodt L，CAＲL Bergmann． Dual energy X-ray transmis-sion sorting of coal［J］．Journal of the Southern African Institute of Mining and Metallurgy，2010，110( 7) : 371－378．

[27]王家臣, 潘卫东, 张国英, 杨胜利, 杨克虎, 李良晖. 图像识别智能放煤技术原理与应用[J]. 煤炭学报, 2022, 47(01):87-101.

[28]任芳. 基于多传感器数据融合技术的煤岩界面识别的理论与方法研究[D]. 太原理工大学, 2003.

[29]王增才, 张秀娟, 张怀新, 王志刚. 自然γ射线方法检测放顶煤开采中的煤矸混合度研究[J]. 传感技术学报, 2003(04):442-446.

[30]冯岸岸. 智能分选过程中煤矸X射线识别技术的研究[D]. 安徽理工大学, 2019.

[31]张新, 王正祥, 武晨光, 王磊. 煤岩界面识别技术研究[J]. 煤矿机械, 2015, 36(02):3-5. DOI:10. 13436/j. mkjx. 201502002.

[32]于中山. 基于X射线图像处理的煤矸识别技术研究[D]. 安徽理工大学, 2020. DOI:10. 26918/d. cnki. ghngc. 2020. 000240.

[33]马昊伟. 基于激光检测技术的干法选煤系统的探索性研究[D]. 陕西科技大学, 2012.

[34]郭永存, 何磊, 刘普壮, 王希. 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. DOI:10. 13225/j. cnki. jccs. 2020. 1626.

[35]Sun J, Su B. Coal–rock interface detection on the basis of image texture features[J]. International Journal of Mining Science and Technology, 2013, 23(5): 681-687.

[36]Liu Cz, Li M, Zhang Y, et al. An enhanced rock mineral recognition method integrating a deep learning model and clustering algorithm[J]. Minerals, 2019, 9(9): 516.

[37]段雍. 基于图像的煤矸识别和定位方法研究与实现[D]. 西安科技大学, 2020. DOI:10. 27397/d. cnki. gxaku. 2020. 000027.

[38]Hou W. Identification of coal and gangue by feed-forward neural network based on data analysis[J]. International Journal of Coal Preparation and Utilization, 2019, 39(1): 33-43.

[39]孙继平, 陈浜. 基于双树复小波域统计建模的煤岩识别方法[J]. 煤炭学报, 2016, 41(07):1847-1858. DOI:10. 13225/j. cnki. jccs. 2015. 1447.

[40]韦任. 基于煤岩反射光谱的识别技术研究[D]. 安徽理工大学, 2021. DOI:10. 26918/d. cnki. ghngc. 2021. 000277.

[41]师瑞卓, 张小俊, 孙凌宇, 裴香丽. 基于单线激光雷达的三维形貌重建方法研究[J]. 激光与红外, 2022, 52(02):188-195.

[42]Cui Yang, Li Qingquan, Yang Bisheng, Xiao Wen, Chen Chi, Dong Zhen. Automatic 3-D Reconstruction of Indoor Environment With Mobile Laser Scanning Point Clouds[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019, 12(8).

[43]Li Shuaihao, He Yanxiang, Li Qianqian, Chen Min. Using Laser Measuring and SFM Algorithm for Fast 3D Reconstruction of Objects[J]. Journal of Russian Laser Research, 2018, 39(6).

[44]乔俊峰, 周沅桢, 王永, 范兴, 农永红, 李博宇, 资文华. 三维激光扫描测体积技术及其应用进展[J]. 激光与红外, 2021, 51(09):1115-1122.

[45]骆钰波, 黄洪宇, 唐丽玉, 陈崇成, 张浩. 基于地面激光雷达点云数据的森林树高、胸径自动提取与三维重建[J]. 遥感技术与应用, 2019, 34(02):243-252.

[46]胡鹏程, 郭焱, 李保国, 朱晋宇, 马韫韬. 基于多视角立体视觉的植株三维重建与精度评估[J]. 农业工程学报, 2015, 31(11):209-214.

[47]吴海滨, 徐若彤, 王爱丽, 于晓洋, 岩堀, 祐之, 赵蓝飞, 刘赫. 基于计算机视觉的人体内腔三维重建技术综述[J]. 计算机工程, 2021, 47(10):1-15.

[48]Hanghang Jiang, Hongxin Zhang, Jiaming Li, Meng Li, Shaowei Ma. Recent Reviews on Machine Vision-Based 3D Reconstruction[J]. Recent Patents on Mechanical Engineering, 2022, 15(1).

[49]Liu Yifan, Cai Zhenjiang. Binocular Stereo Vision Three-Dimensional Reconstruction Algorithm Based on ICP and SFM[J]. Laser & Optoelectronics Progress, 2018, 55(9).

[50]Lee Kyungchai, Lee Onseok. A Study on the Optimization for Three Dimensional Reconstruction of Bio Surface Using by Stereo Vision[J]. The Transactions of The Korean Institute of Electrical Engineers, 2017, 66(1).

[51]Patel D. K. , Bachani P. A. , Shah N R. Distance measurement system using binocular stereo vision approach[J]. International Journal of Engineering Research &Technology, 2013, 2(12): 2461-2464.

[52]Barnard S T, Thompson W B. Disparity analysis of images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1980 (4): 333-340.

[53]U. R. Dhond, J. K. Aggarval. Struct from Stereo—A Review [J]. IEEE Trans. Systems, Man, and Cybemeties. 1989, Vol. 19: 1489-1510.

[54]Tomasi C, Kanade T. Shape and motion from image streams under orthography: a factorization method[J]. International journal of computer vision, 1992, 9(2): 137-154.

[55]Koch R, Pollefeys M, Gool L J V. Realistic surface reconstruction of 3D scenes from uncalibrated image sequences[J]. The Journal of Visualization and Computer Animation, 2000, 11(3):115-127.

[56]N. Snavely, S. M. Seitz, R. Szeliski. Modeling the World from Internet Photo Collections[J]. International Journal of Computer Vision, 2008, 80(2):189-210.

[57]刘一凡, 蔡振江. 基于ICP与SFM的双目立体视觉三维重构算法[J]. 激光与光电子学进展, 2018, 55(09):309-317.

[58]魏耀. 基于随机搜索和准稠密匹配的多视角立体重建算法研究[D]. 西安市:西安电子科技大学, 2018.

[59]岳志强. 基于CNN的双目立体视觉三维实时重建算法研究[D]. 沈阳市:沈阳工业大学, 2019.

[60]刘立强. 基于深度学习的视觉三维重建研究[D]. 北京市:北方工业大学, 2019.

[61]李志鹏. 基于双目立体视觉的三维重建[D]. 河北地质大学, 2022.

[62]陈杰, 林晓青, 陆胜勇, 李晓东, 严建华. 基于双目视觉耦合激光的垃圾焚烧炉进料速率实时测量技术[J]. 环境工程学报, 2021, 15(10):3316-3323.

[63]梁乐. 基于双目立体视觉的不规则物体体积测量方法研究[D]. 西安理工大学, 2019.

[64]范徐萌. 基于双目立体视觉的物体体积测量研究[D]. 中国矿业大学, 2021. DOI:10. 27623/d. cnki. gzkyu. 2021. 000394.

[65]刘九卿. 动态和数字称重技术发展概况与研究课题[J]. 工业计量, 2011, 21(02):4-10.

[66]刘九卿. 应变式称重传感器及其相关技术的新发展[J]. 衡器, 2010, 39(11):3-10.

[67]韩昱, 李丽．动态称重系统及数据处理算法的研究[J]．电脑开发与应用, 2006, 19(6)：31-33.

[68]刘悦. 基于PLC控制的取料机在线动态称重计量[D]. 燕山大学, 2017.

[69]李萍, 任安祥. 基于机器视觉的带送煤炭体积测量方法研究[J]. 工矿自动化, 2018, 44(0 4):24-29. DOI:10. 13272/j. issn. 1671-251x. 2017110027.

[70]高艳雯．输送带在线动态实时称重系统的研究[D]．兰州：兰州理工大学, 2007.

[71]李莹. 基于深度学习的煤矸石目标检测方法研究[D]. 西安科技大学, 2020. DOI:10. 27397/d. cnki. gxaku. 2020. 000223.

[72]高飞. Z56151工作面煤和矸石的密度及煤矸比确定[J]. 煤炭与化工, 2017, 40(01):54-56. DOI:10. 19286/j. cnki. cci. 2017. 01. 015.

[73]王红菊, 王永春, 姜志刚. 元宝山露天矿区元宝山组沉积环境及聚煤规律[J]. 科学技术与工程, 2007(10):2326-2330.

[74]王伟. 大柳塔选煤厂入选原煤硫分分布规律研究[J]. 洁净煤技术, 2020, 26(06):83-88. DOI:10. 13226/j. issn. 1006-6772. 20090701.

[75]Li P, Wang Y W, Xu C. Research on dynamic measurement system for bulk material based on machine vision[C]//Applied Mechanics and Materials. Trans Tech Publications Ltd, 2013, 273: 768-772.

[76]曾灿灿, 任明俊, 肖高博, 殷跃红. 基于贝叶斯推理的多尺度双目匹配方法[J]. 光学学报, 2017, 37(12):263-271.

[77]He Tao, Chen Jiuyin, Hu Xiang, et al. A Study of 3d Coordinate Measuring Based on Binocular Stereo Vision[J]. Applied Mechanics & Materials, 2015, 740: 531-534.

[78]黄风山, 刘恩福, 方忆湘, 韩宁. 基于智能三坐标测量机的零件位姿单目立体视觉识别[J]. 光学精密工程, 2013, 21(05):1326-1332.

[79]叶道焕, 黄庚华, 程鹏飞. 高精度激光动态测试技术研究[J]. 红外, 2016, 37(05):7-9+16.

[80]胡伟伟, 李永亮, 顾小琨, 张翼鹏, 张英明, 刘泓鑫. 远程激光测距技术及其进展[J]. 激光与红外, 2019, 49(03):273-281.

[81]范徐萌. 基于双目立体视觉的物体体积测量研究[D]. 中国矿业大学, 2021.

[82]张晨. 煤矸光电密度识别及自动分选系统的研究[D]. 中国矿业大学(北京), 2013.

[83]Lowe D G. Distinctive image features from scale-invariant keypoints[J]. International journal of computer vision, 2004, 60(2): 91-110.

[84]王鹏, 曹现刚, 马宏伟, 吴旭东, 夏晶. 基于余弦定理-PID的煤矸石分拣机器人动态目标稳准抓取算法[J]. 煤炭学报, 2020, 45(12):4240-4247. DOI:10. 13225/j. cnki. jccs. 2019. 1565.