随着煤矿智能化水平的提升，煤矿智能化综采工作面普遍采用视频监控技术进行远程监测。在综采工作面生产过程中，单个摄像头视野有限，无法实现大范围监控，难以满足人工视频远程干预需求。综采工作面的安全、高效、智能生产，需要对综采工作面刮板输送机直线度进行检测，为液压支架自动调直奠定基础。本文采用机器视觉方法，研究综采工作面视频图像拼接及刮板输送机直线度检测方法，对煤矿综采工作面智能开采具有重要意义。

针对综采工作面雾尘干扰导致的图像质量差问题，建立了基于引导滤波的雾尘图像清晰化模型，该模型通过引入引导滤波对透射率函数进行优化以提高图像质量。与多种目前广泛使用的图像清晰化算法进行对比，结果表明：本文提出的方法在视觉上和客观评价指标上都具有最好的效果，适用于综采工作面环境，为后续综采工作面视频拼接和刮板输送机直线度检测提供良好基础。

针对综采工作面恶劣环境视频图像拼接的特征提取与匹配难题，提出了一种改进SURF-FLANN的综采工作面视频拼接特征提取与匹配算法。该方法在特征提取方面，通过将高斯滤波改为双边滤波提取图像中的SURF关键特征点，并提出改进描述符的方法解决综采工作面视频图像特征点误提取问题。在特征匹配方面，该方法以FLANN算法为基础，结合随机抽样一致（RANSAC）算法思想，减少大量误匹配、无用的特征点，解决原FLANN算法匹配速度和正确率低的问题。通过多种匹配算法对综采工作面视频图像进行特征提取与匹配对比实验，结果表明：提出的方法在客观评价指标上都具有良好的效果，特征提取与匹配的平均匹配正确率和平均匹配速度最高，分别达到81.47%和51.5fps，适用于综采工作面环境，为后续综采工作面视频拼接奠定良好基础。

针对综采工作面视频拼接时摄像头拍摄画面存在图像畸变和融合图像出现重影、错位等问题，提出一种最佳缝合线拼接算法。该方法通过对图像进行畸变矫正，减少噪声干扰以及降低失真度。在重叠区域中选择能量差异函数最小位置的一条缝合线，根据缝合线位置剪切图像，并将剪切后的图像投影到同一幅图像上，解决融合图像出现重影、错位的问题。通过多种图像拼接算法进行对比实验，结果表明：提出的方法在视觉效果和客观评价指标均优于其他常用视频图像融合算法，且双摄像头、四摄像头、实际综采工作面视频的拼接速度分别达到了40.2fps、22.04fps、38.7fps。为后续刮板运输机直线度检测提供了高质量的图像。

针对刮板输送机直线度测量的问题，提出一种基于机器视觉与改进LSM的刮板输送机直线检测算法，该算法采用EDLines算法进行直线检测，通过标定兴趣区检测刮板运输机边缘直线和合并直线并进行曲线拟合来判定刮板输送机直线度误差。通过多种刮板运输机直线度检测方法进行对比实验，结果表明：本文提出的方法对15m长的刮板运输机的直线状态误差和水平弯曲状态误差小于3mm，以及检测速度为0.122s。

通过煤矿综采工作面的监控视频对本文提出的视频拼接和刮板输送机直线度检测方法进行了实验验证，结果表明：综采工作面视频拼接方法中的特征提取与匹配的平均匹配正确率和平均匹配速度最高，平均正确率为81.47%和平均匹配速度为51.5fps，且拼接效果都优于其他算法。此外，将提出的最佳缝合线融合方法和其他图像融合算法进行实验对比，结果表明：提出的最佳缝合线融合方法在图像拼接部分过度更加平滑，拼接速率达到38.7fps。提出的刮板运输机直线度检测方法可以精确检测出刮板运输机直线度且误差最小，直线状态误差和水平弯曲状态误小于3mm，并且检测速度达到了0.122s。能够实现综采工作面刮板运输机直线度精准、实时检测。

With the improvement of the intelligent level of coal mine, the intelligent fully-mechanized mining face generally uses video surveillance technology for remote monitoring. In the production process of fully mechanized mining face, the field of view of a single camera is limited, which can not realize large-scale monitoring, and it is difficult to meet the needs of remote intervention by artificial video. The safety, high efficiency and intelligent production of fully mechanized mining face need to detect the straightness of the scraper conveyor and lay the foundation for the automatic straightening of the hydraulic support. In this paper, the machine vision method is used to study the video image splicing and the straightness detection method of scraper conveyor, which is of great significance to the intelligent mining of fully mechanized mining face.

Aiming at the problem of poor image quality caused by fog and dust interference in fully mechanized mining face, a clear model of fog and dust image is established based on guided filter, which optimizes transmission function by introducing guided filter to improve image quality. Compared with a variety of widely used image sharpening algorithms, the results show that the method proposed in this paper has the best effect in terms of vision and objective evaluation indicators, and is suitable for the fully mechanized mining face environment, providing a good basis for the subsequent video splicing of fully mechanized mining face and the straightness detection of scraper conveyor.

Aiming at the problem of feature extraction and matching of video image Mosaic in harsh environment of fully mechanized mining face, an improved SURF-FLANN algorithm for feature extraction and matching of video Mosaic of fully mechanized mining face is proposed. In the aspect of feature extraction, the SURF key feature points in images were extracted by changing the Gaussian filter to bilateral filter, and an improved descriptor method was proposed to solve the problem of misextraction of feature points in fully mechanized mining face video images. In the aspect of feature matching, this method is based on FLANN algorithm and combined with RANSAC algorithm, which reduces a large number of mismatched and useless feature points, and solves the problem of low matching speed and accuracy of the original FLANN algorithm. Through a variety of matching algorithms, feature extraction and matching comparison experiments are carried out on video images of fully mechanized mining face. The results show that: The proposed method has good results in the objective evaluation indicators, and the average matching accuracy and average matching speed of feature extraction and matching are the highest, reaching 81.47% and 51.5fps respectively, which is suitable for the environment of fully mechanized mining face and lays a good foundation for the follow-up video Mosaic of fully mechanized mining face.

Aiming at the problem of image distortion, double shadow and dislocation in the fused images, an optimal stitching algorithm was proposed. The method can reduce the noise interference and the distortion degree by correcting the image distortion. A suture line with the smallest energy difference function is selected in the overlapping region, and the image is cut according to the position of the suture line, and the cut image is projected onto the same image to solve the problem of double shadow and dislocation in the fused image. The results show that the proposed method is superior to other common video image fusion algorithms in terms of visual effects and objective evaluation indexes, and the video fusion speed of dual-camera, four-camera and actual fully-mechanized mining face reaches 40.2fps, 22.04fps and 38.7fps, respectively. High quality images are provided for the subsequent detection of the straightness of the scraper conveyer.

Aiming at the problem of measuring the straightness of the scraper conveyor, a straight-line detection algorithm based on machine vision and improved LSM is proposed. The algorithm adopts the EDLines algorithm for straight-line detection, and determines the straight-line error of the scraper conveyor by calibrating the interest zone to detect the edge straight line and the combined straight line, and carries out curve fitting. The results show that the linear state error and horizontal bending state error of 15m long scraper conveyer are less than 3mm, and the detection speed is 0.122s.

The video splicing and scraper conveyor straightness detection methods proposed in this paper are experimentally verified through the monitoring video of the fully-mechanized mining face in coal mine. The results show that: The average matching accuracy and average matching speed of feature extraction and matching method in fully mechanized mining face video stitching are the highest, the average accuracy is 81.47% and the average matching speed is 51.5fps, and the stitching effect is better than other algorithms. In addition, the experimental comparison between the proposed optimal suture fusion method and other image fusion algorithms shows that the proposed optimal suture fusion method is smoother in the image stitching part, and the stitching rate reaches 38.7fps. The proposed method can accurately detect the straightness of the scraper conveyor with the smallest error, the linear state error and the horizontal bending state error are less than 3mm, and the detection speed reaches 0.122s. It can realize accurate and real-time detection of straightness of scraper conveyer on fully mechanized mining face.

[1]李霞,崔涛. 我国煤炭资源可持续发展的保障分析[J]. 中国煤炭, 2019, 45(01): 33-37.

[2]张莹. 我国煤炭转型面临的挑战与对策[J]. 环境保护, 2018, 46(02): 24-29.

[3]钱鸣高,许家林,王家臣. 再论煤炭的科学开采[J]. 煤炭学报, 2018, (01): 1-13.

[4]En Lu, Wei Li, Xuefeng Yang, et al.Composite sliding mode control of a permanent magnet direct-driven system for mining scraper conveyor[J]. IEEE Access, 2017, 5.

[5]牛剑峰. 综采工作面直线度控制系统研究[J].工矿自动化,2015,41(5):5-8.

[6]Einicke G, Ralston J, Hargrave C, et al.Longwall mining automation an application of minimum-variance smoothing [J].IEEE Control Systems Magazine,2008,28(6):28-37.

[7]Wang G, Jiao S, Cheng G.Fully mechanized coal mining technology for thin coal seam under complicated geological conditions[J].Energy Exploration & Exploitation, 2011, 29(2):169-178.

[8]罗成名. 链式传感网中煤矿井下移动装备位姿感知理论及技术研究[D].中国矿业大学,2014.

[9]樊启高. 综采工作面“三机”控制中设备定位及任务协调研究[D].中国矿业大学,2013.

[10]王崟. 基于景深提取和区域分割的图像去雾算法优化与实现[D].南京航空航天大学,2018.

[11]孟宇飞,王晓玲,刘畅,周哲海.四分暗通道均值比较法的双角度偏振图像去雾[J].激光与光电子学进展,2022,59(04):240-248.

[12]刘玲. 基于暗通道先验的图像去雾改进算法研究[D].重庆大学,2019.

[13]He Kai-ming，Sun Jian，Tang Xiao-ou．Single image haze removal using dark channel prior［J］．IEEE Transactions on Pattern Analysis and M achine Intelligence，2011,33(12):2341-2353．

[14]He K M, Sun J, Tang X O. Guided image filtering[J]. IEEE Transactions on Software Engineering, 2013, 35(6):1397-1409.

[15]Park D, Han D K, Ko H. Single image haze removal with WLS-based edge-preserving smoothing filter[C]. IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2013:2469-2473.

[16]Gibson K B, Nguyen T Q. Fast single image fog removal using the adaptive Wiener filter[C]. IEEE International Conference on Image Processing. IEEE, 2014:714-718.

[17]Yu J, Xiao C, Li D. Physics-based fast single image fog removal[C]. IEEE, International Conference on Signal Processing. IEEE, 2010:1048-1052.

[18]Xiao C, Gan J. Gan, J.: Fast image dehazing using guided joint bilateral filter. Vis. Comput. 28(6-8), 713-721[J]. Visual Computer, 2012, 28(6-8):713-721.

[19]曹虎晨,姚善化,王仲根. 基于边界约束的煤矿井下尘雾图像去雾算法[J].工矿自动化,2022,48(06):139-146.

[20]王媛彬,韦思雄,段誉,吴华英.基于自适应双通道先验的煤矿井下图像去雾算法[J].工矿自动化,2022,48(05):46-51+84.

[21]王启明,李季.煤矿井下高清图像快速去雾算法研究[J].小型微型计算机系统,2018,39(11):2557-2560.

[22]冯卫兵,胡俊梅,曹根牛.基于改进的简化脉冲耦合神经网络的煤矿井下图像去噪方法[J].工矿自动化,2014,40(05):54-58.

[23]Mao Q, Wang Y, Zhang X, et al. Clarity method of fog and dust image in fully mechanized mining face[J]. Machine Vision and Applications, 2022, 33(2): 1-16.

[24]付翘楚.视频图像拼接技术研究[D]. 北京: 北京邮电大学, 2017.

[25]程德强,厉航,黄晓丽等. 基于改进随机抽样一致算法的视频拼接算法[J]. 工矿自动化, 2017, 43(08): 50-55.

[26]管增伦,顾军,赵广源.基于改进加速鲁棒特征的综采工作面视频拼接算法[J]. 工矿自动化, 2018, 44(11): 69-74.

[27]高尚,董爽,孙文雪等.基于特征点聚类的立井井壁图像匹配算法[J].信息技术与信息化,2021(03):73-75.

[28]白雪冰,车进,吴金蔓等.基于Transformer视觉特征融合的图像描述方法[J/OL].计算机工程:1-12.

[29]王红尧,吴佳奇,林松等.矿井多视角图像拼接方法研究[J].工矿自动化,2021,47(10):27-31+48.

[30]杨炳坤,程树英,郑茜颖. 一种面向图像拼接的改进 PCA-SIFT 算法[J]. 微电子学与计算机, 2018, 35(12): 70-75.

[31]智宁, 毛善君, 李梅, 等. 基于深度融合网络的煤矿图像尘雾清晰化算法[J]. 煤炭学报, 2019, 44(02): 655-666.

[32]马宏伟,张烨,王鹏,等. 多机械臂煤矸石智能分拣机器人关键共性技术研究[J]. 煤炭科学技术,2023,51(1):427-436.

[33]潘峰,沈建新,秦顺等.融合GMS的ORB特征点提取与匹配算法[J].计算机工程与设计,2022,43(08):2244-2251.

[34]柴成富. 基于图像识别的矿山相似材料实验模型变形信息提取[D].西安科技大学,2022.

[35]周磊,华钢,徐钊等.基于连续尺度空间和多特征提取的煤矿监控视频注意计算模型[J].煤炭学报,2010,35(08):1406-1411.

[36]姜丽萍,熊淑华,员旭拓等.基于SR-SURF的岩石显微图像拼接[J/OL].计算机系统应用:1-6[2023-10-16].

[37]冯宝凤,杨剑锋,严可等.融合RANSAC与GMS图像特征的匹配算法[J].智能计算机与应用,2023,13(09):74-79.

[38]Ying Z ,Hongliang G ,Fuzhou D , et al.A Light Field Full-Focus Image Feature Point Matching Method with an Improved ORB Algorithm[J].Sensors,2022,23(1):123-123.

[39]Choi K, Jun K. Real-time panorama video system using networked multiple cameras[J]. Journal of Systems Architecture, 2016, 64: 110-121.

[40]Alomran M, Chai M. Feature-based panoramic image stitching[C]// 14th International Conference on Control, Automation, Robotics and Vision. Phuket: 2016: 1-6.

[41]V M M Babu, Santha T. Efficient brightness adaptive deep-sea image stitching using biorthogonal multi-wavelet transform and harris algorithm[C]// 2017 International Conference on Intelligent Computing and Control. Coimbatore: 2017: 1-5.

[42]Hejazifar H, Khotanlou H. Fast and robust seam estimation to seamless image stitching[J]. Signal Image and Video Processing. 2018, 12(5): 885-893.

[43]Widiyaningtyas T, D.D.Prasetya, A.P.Wibawa. Web-based Campus Virtual Tour Application usingORB Image Stitching[C]// 5th International Conference on Electrical Engineering, Computer Science and Informatics, IEEE, Malang, Indonesia, 2018:46-49.

[44]Ahn H, Lee JH, Cho HJ. Research of Panoramic Image Generation Using IoT Device with Camera for Cloud Computing Environment[J]. Wireless Personal Communications. 2019, 105(2): 619-634.

[45]Laaroussi S, Baataoui A, Halli A, et al. Dynamic mosaicking: region-based method using edgedetection for an optimal seamline[J]. Multimedia Tools and Applications, 2019, 78(16): 23225-23253.

[46]Kim BS, Choi KA, Park WJ, et al. Content-Preserving Video Stitching Method for Multi-Camera Systems[J]. IEEE Transactions on Consumer Electronics, 2017, 63(2): 109-116.

[47]Yeh H, Lai S. Real-time video stitching[C]. 2017 IEEE International Conference on Image Processing (ICIP), Beijing, 2017: 1482-1486.

[48]张帆,李闯,李昊.智能综采工作面刮板输送机直线度监测方法研究[J].煤炭科学技术,2022,50(04):246-255.

[49]夏婷. 综采工作面刮板输送机直线度检测方法研究[D].中国矿业大学,2019.

[50]刘鹏坤. 基于视觉测量的综采工作面直线度控制研究[D].中国矿业大学(北京),2020.

[51]王雪松,王世博,王世佳,李争,葛世荣.刮板输送机直线度误差预测模型[J/OL].中国矿业大学学报:1-10.

[52]石金龙. 惯导与里程计组合的综采工作面直线度测量技术研究[D].西安科技大学,2021.

[53]杨曌.基于视觉测量的综采工作面直线度测量技术研究[J].煤炭工程,2016,48(09):134-136.

[54]余佳鑫,煤矿综采工作面可视化系统的研究与实现.广东省,日立楼宇技术,2020-03-20.

[55]牛剑峰.基于视频巡检的综采工作面无人化关键技术研究[J].煤炭科学技术,2019,47(10):141-146.

[56]葛世荣,王俊涛,宋智丽.刮板输送机技术发展历程(一)国外技术[J].中国煤炭,2024,50(02):1-12.

[57]Meng G, Wang Y, Duan J, et al. Efficient image dehazing with boundary constraint and contextual regularization[C]. Proceedings of the IEEE international conference on computer vision. 2013: 617-624.

[58]Galdran A. Image dehazing by artificial multiple-exposure image fusion[J]. Signal Processing, 2018, 149: 135-147.

[59]Bay H ,Ess A ,Tuytelaars T ,et al.Speeded-Up Robust Features (SURF)[J].Computer Vision andImage Understanding,2007,110(3):346-359.

[60]Zhang H ,Zheng R ,Zhang W ,et al.An Improved SIFT Underwater Image Stitching Method[J].Applied Sciences,2023.

[61]高槐泽.智能电铲三维环境感知与挖掘轨迹规划研究[D].大连理工大学,2022.

[62]曹媛.基于哈希学习的近似最近邻搜索方法的研究[D].大连理工大学,2019.

[63]P.H.S. T ,C. D .IMPSAC: synthesis of importance sampling and random sample consensus[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2003,25(3):354-364.

[64]罗永涛,王艳,张红民.结合最佳缝合线和改进渐入渐出法的图像拼接算法[J].红外技术,2018,40(04):382-387.

[65]喻富. 复杂背景下基于全景图像拼接的人脸检测算法研究[D].湖南大学,2014.

[66]么鸿原,林雪原,王海鹏等.针对高分辨率遥感图像的融合拼接算法[J].计算机工程与设计,2019,40(11):3230-3235.

[67]赵光远.基于改进APAP的图像配准和最佳缝合线融合的图像拼接研究[D].西安理工大学,2023.

[68]吴雨菲.基于深度特征和最佳缝合线的图像拼接方法研究[D].中国矿业大学,2023.

[69]Akinlar C ,Topal C .EDLines: A real-time line segment detector with a false detection control[J].Pattern Recognition Letters,2011,32(13):1633-1642.

[70]Rafael Grompone von Gioi, Jérémie Jakubowicz, Jean-Michel Morel, Gregory Randall. LSD: a Line Segment Detector[J]. Image Processing On Line, 2012(2): 35-55.

[71]陈桂秀. 用程序求解最小二乘拟合多项式的系数[J]. 青海师范大学学报(自然科学版), 2010, 26(03): 14-17.

[72]陆健. 最小二乘法及其应用[J]. 中国西部科技, 2007(19): 19-21

[73]高秋英,王丽丽,王荣忠.最小二乘法曲线拟合及优化算法研究[J].工业控制计算机,2021,34(11):100-101.

[74]王一函,凌鹤.基于响应面的直线导轨性能参数与矫直行程预测模型仿真[J].数字制造科学,2023,21(04):248-252.

[75]王可,宋天皎,孙兴伟等.基于最小区域法的平面内直线度误差评定方法可视化设计[J].机械,2016,43(03):6-8+19.

[76]左倩.用最小区域法求直线度误差的探讨[J].科技创新与应用,2015,(23):84-85.

[77]Galdran A. Image dehazing by artificial multiple-exposure image fusion[J]. Signal Processing, 2018, 149: 135-147.

[78]Ehsan S M, Imran M, Ullah A, et al. A Single Image Dehazing Technique Using the Dual Transmission Maps Strategy and Gradient-Domain Guided Image Filtering[J]. IEEE Access, 2021,9: 89055-89063.

[79]Berman D, Treibitz T, Avidan S. Air-light estimation using haze-lines[C]. 2017 IEEE Internati-onal Conference on Computational Photography (ICCP). IEEE, 2017: 1-9.

[80]Shangyu Du, Hongyu Shi. Development of optical fiber vector sensor and its application in straightness monitoring of scraper conveyor[J].02 February 2024.

[81]Benyong C ,Enzheng Z ,Liping Y , et al.A laser interferometer for measuring straightness and its position based on heterodyne interferometry.[J].The Review of scientific instruments,2009,80(11):115113.

[82]Wang B ,Xie J ,Wang X , et al.A New Method for Measuring the Attitude and Straightness of Hydraulic Support Groups Based on Point Clouds[J].Arabian Journal for Science and Engineering,2021,46(12):1-19.

[83]Jiakun L ,Peng S ,Zhijun W , et al.Research on a Real-Time Monitoring Method for the ThreeDimensional Straightness of a Scraper Conveyor Based on Binocular Vision[J].Mathematics,2022,10(19):3545-3545.

[84]Wang B ,Xie J ,Wang X , et al.A New Method for Measuring the Attitude and Straightness of Hydraulic Support Groups Based on Point Clouds[J].Arabian Journal for Science and Engineering,2021,46(12):1-19.

[85]王超,李威,杨海,等.基于航位推测的刮板输送机形态检测研究[J].煤炭学报,2017, 42(8):2173-2180.