煤矿掘进安全是掘进工作面智能化建设的重要内容，近年来，为了提高掘进效率，许多煤矿引进了各种大功率掘进装备，但由于大型掘进机体积庞大且工作时产生的噪声大，当矿工无意识进入掘进机周边危险区域时，会出现监管人员被遮挡视角和听不见人员呼喊等情况，这将造成很大的危险隐患或出现重大安全事故。因此，本文对掘进工作面危险区域人员入侵预警方法进行研究，构建掘进工作面危险区域坐标模型，提出基于激光雷达与相机融合的方法对危险区域人员进行定位与识别，并将人员坐标与危险区域坐标结合进行阈值判断，构建人员定位与预警的可视化界面，实现人员的入侵预警。论文主要研究工作包括：

针对激光雷达原始点云数据量庞大、包含大量无效点云且大量地面点云影响后续聚类效果等问题，本文采用体素滤波和统计滤波算法去除原始点云噪声并降低原始点云数量，并提出一种改进的RANSAC地面点云分割算法，在原始RANSAC算法上引入法向量校验和高度限制，解决误拟合平面的问题。对处理后的离散点云进行欧式聚类，将聚类后的点云簇拟合成AABB包围盒，并计算拟合包围盒的中心坐标作为人员的三维坐标。

针对激光雷达点云数据缺乏语义信息导致无法分辨目标类别的问题，提出一种基于单目视觉的YOLOv7人员目标识别方法。建立相机模型，通过棋盘格标定法，完成相机的内参标定和畸变矫正。然后对相机采集的原始图像进行基于改进Retinex算法的图像增强处理。自制掘进工作面数据集，通过YOLOv7网络进行人员目标检测训练，完成人员目标识别。

研究激光雷达人员定位信息与单目相机人员识别信息融合技术，提出基于人员定位融合信息的掘进工作面危险区域入侵方法。对相机和激光雷达进行联合标定与时间同步，得到激光雷达坐标系与像素坐标系转换矩阵，通过转换矩阵将点云聚类结果投影到二维相机平面，得到二维点云检测框，然后，采用基于DIoU融合算法融合相机与激光雷达检测框，输出检测类型与坐标，完成人员定位与识别任务。构建掘进机危险区域二维坐标模型，根据人员定位得到的x、y方向的坐标数据，结合危险区域坐标进行阈值判断，完成入侵预警。搭建掘进工作面危险区域人员入侵预警系统并设置预警效果。

最后，模拟掘进工作面环境，搭建实验平台，对人员定位识别方法和人员入侵预警方法进行功能验证与性能测试。实验结果表明，基于激光雷达与相机的人员定位识别方法的位置平均误差不超过50mm。人员的检测准确率为96.4%，召回率为97.1%，推理速度为96.2ms，有效保证了煤矿井下复杂、多变环境中人员入侵预警功能的鲁棒性，对促进智能化掘进技术发展具有重大意义。

Coal mining safety is an important aspect of intelligent construction in coal mining operations. In recent years, various high-power excavation equipment has been introduced into many coal mines to improve excavation efficiency. However, due to the large size and significant noise generated during operation of large-scale excavation machinery, there is a risk of obscured visibility and inability to hear calls from workers when they inadvertently enter hazardous areas around the machinery. This poses significant safety hazards and the potential for major accidents. Therefore, this paper studies the method of personnel intrusion warning in hazardous areas of the heading, constructs a coordinate model of hazardous areas, proposes a method based on the fusion of LiDAR and cameras to locate and identify personnel in hazardous areas, and combines personnel coordinates with hazardous area coordinates for threshold judgment. A visualization interface for personnel positioning and warning is developed to achieve intrusion warning. The main research work of the paper includes:

To address the challenges of large volumes of raw LiDAR point cloud data, which often contain numerous invalid points and significant ground points that may affect subsequent clustering, this paper proposes the use of voxel and statistical filtering algorithms to remove noise and reduce the volume of the original point cloud data. Additionally, an improved RANSAC ground point segmentation algorithm is introduced, which incorporates normal vector validation and height constraints to address the issue of misfitting planes encountered in the original RANSAC algorithm. Subsequently, Euclidean clustering is applied to the processed point cloud to group points into clusters. These clustered point cloud clusters are then fitted with axis-aligned bounding boxes (AABB), and the center coordinates of the fitted bounding boxes are calculated as the three-dimensional coordinates of personnel.

To address the lack of semantic information in LiDAR point cloud data, which hinders the ability to discern target categories, this paper proposes a YOLOv7-based personnel object recognition method utilizing monocular vision. Firstly, a camera model is established, and camera intrinsic calibration and distortion correction are achieved using a chessboard calibration method. Subsequently, the original images captured by the camera are subjected to image enhancement using an improved Retinex algorithm. A custom dataset of heading scenarios is created, and personnel object detection training is conducted using the YOLOv7 network to achieve personnel object recognition.

Research on the fusion technology of personnel positioning information from LiDAR and personnel recognition information from monocular cameras, proposing an intrusion method for hazardous areas in headings based on fused personnel positioning information. Firstly, joint calibration and time synchronization are conducted for the camera and LiDAR to obtain the transformation matrix between the LiDAR coordinate system and the pixel coordinate system. Using this transformation matrix, the clustered point cloud results are projected onto the two-dimensional camera plane to obtain two-dimensional point cloud detection boxes. Then, a fusion algorithm based on Distance-IoU (DIoU) is applied to merge the camera and LiDAR detection boxes, outputting the detection types and coordinates to complete the personnel positioning and recognition tasks. A two-dimensional coordinate model of the hazardous area in the heading is constructed. Threshold judgment is conducted based on the x and y coordinates obtained from personnel positioning, combined with the coordinates of the hazardous area, to complete intrusion warning. Finally, a personnel intrusion warning system for hazardous areas in headings is established and its warning effects are configured.

In conclusion, a simulated heading environment was created, and an experimental platform was established to conduct functional verification and performance testing of the personnel positioning and recognition methods, as well as the personnel intrusion warning method. The experimental results show that the average position error of the personnel positioning and recognition method based on LiDAR and cameras does not exceed 50mm. The detection accuracy of personnel is 96.4%, with a recall rate of 97.1%, and an inference speed of 96.2ms. This effectively ensures the robustness of personnel intrusion warning in the complex and dynamic underground environment of coal mines, which is of significant importance for promoting the development of intelligent excavation technology.

[1] 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.

[2] 王国法,刘峰,庞义辉,等.煤矿智能化——煤炭工业高质量发展的核心技术支撑[J].煤炭学报,2019,44(02):349-357.

[3] 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.

[4] 王步康.煤矿巷道掘进技术与装备的现状及趋势分析[J].煤炭科学技术,2020,48(11):1-11.

[5] 国家安全生产监督管理总局.煤矿安全规程[M].北京:应急管理出版社,2022.

[6] 倪云峰,王志刚,王静,等.基于RSSI的井下人员定位算法改进[J].无线电工程,2023,53(03):663-668.

[7] 惠威威,肖广兵.基于GPS定位的新冠防疫辅助系统[J].农业装备与车辆工程,2020,58(11):67-70.

[8] Bai Y B, Wu S, Wu H R, et al. Overview of RFID-Based Indoor Positioning Technology[J]. GSR, 2012, 2012.

[9] Cheng Y K, Chou H J, Chang R Y. Machine-learning indoor localization with access point selection and signal strength reconstruction[C]//2016 IEEE 83rd Vehicular Technology Conference (VTC Spring). IEEE, 2016: 1-5.

[10] Dang X, Zhang T, Hao Z. WiFi-based indoor personnel location monitoring method[C]//IOP Conference Series: Materials Science and Engineering. IOP Publishing, 2019, 569(3): 032076.

[11] 王鑫,王向军.大视场双目视觉定位系统中多目标稀疏匹配[J].红外与激光工程,2018,47(07):302-307.

[12] 魏军,韩君龙,王巍,等.基于毫米波雷达的铁路隧道人员定位技术研究[J].隧道建设(中英文),2023,43(S2):223-229.

[13] Ibisch A, Stümper S, Altinger H, et al. Towards autonomous driving in a parking garage: Vehicle localization and tracking using environment-embedded lidar sensors[C]//2013 IEEE intelligent vehicles symposium (IV). IEEE, 2013: 829-834.

[14] 齐振磊,付天光,张琨.露天煤矿车辆和人员定位管理系统的设计与应用[J].煤炭工程,2019,51(S1):12-16.

[15] 周海峰,何勇.基于物联网技术的养老院人员安全健康监控系统设计[J].软件工程,2022,25(05):19-22.

[16] 蔡安江,徐海涛,程东波,等.矿井多人员定位轨迹的预警分类方法研究[J/OL].金属矿山,1-10[2024-04-24].

[17] Liu F, Li X, Wang J, et al. An adaptive UWB/MEMS-IMU complementary kalman filter for indoor location in NLOS environment[J]. Remote sensing, 2019, 11(22): 2628.

[18] 刘晨旭,王兴众,郭浩年.基于WIFI指纹的船舶人员定位算法[J].船舶物资与市场,2023,31(08):106-111.

[19] 郭瑜.基于智能手机和ZigBee技术的煤矿人员定位研究[J].煤炭技术,2023,42(11):237-240.

[20] 张华刚.基于机器视觉算法的车站人员定位方法[J].自动化与仪器仪表,2023(02):81-86.

[21] 魏林.基于目标检测与双目视觉的目标定位算法研究与应用[D].重庆大学,2021.

[22] Zhai G, Zhang W, Hu W, et al. Coal mine rescue robots based on binocular vision: A review of the state of the art[J]. IEEE Access, 2020, 8: 130561-130575.

[23] Wu C, Zhang F, Wang B, et al. mmTrack: Passive multi-person localization using commodity millimeter wave radio[C]//IEEE INFOCOM 2020-IEEE Conference on Computer Communications. IEEE, 2020: 2400-2409.

[24] 张倩榕,李毅.NLOS环境下多传感器融合的室内定位方法[J].计算机工程与设计,2023,44(03):732-738.

[25] Ciuffreda I, Casaccia S, Revel G M. A Multi-Sensor Fusion Approach Based on PIR and Ultrasonic Sensors Installed on a Robot to Localise People in Indoor Environments[J]. Sensors, 2023, 23(15): 6963.

[26] 王随阳,蒋林,雷斌,等.基于多传感器融合的室内机器人定位研究[J].组合机床与自动化加工技术,2020,(10):72-75+89.

[27] 孙伟,孙沛伦.非视距环境下的UWB/INS室内定位方法[J].测绘科学,2023,48(01):1-7.

[28] 颜俊杰,蔡芸,蒋林,等.基于EKF的多传感器融合定位算法研究[J].农业装备与车辆工程,2024,62(03):91-95.

[29] 徐淑萍,杨定哲,熊小墩.多传感器融合的室内机器人SLAM[J].西安工业大学学报,2024,44(01):93-103.DOI:10.16185/j.jxatu.edu.cn.2024.01.401.

[30] Yuan C, Liu X, Hong X, et al. Pixel-level extrinsic self calibration of high resolution lidar and camera in targetless environments[J]. IEEE Robotics and Automation Letters, 2021, 6(4): 7517-7524.

[31] 李玉刚,陈克,董振飞.基于多传感器融合的目标检测方法研究[J].内燃机与配件,2024,(07):31-33.

[32] 姬鹏,高帅轩.基于ROS机器人的激光与视觉融合建图方法[J].组合机床与自动化加工技术,2022,(12):132-135.

[33] 黄远宪,李必军,黄琦,等.融合相机与激光雷达的目标检测、跟踪与预测[J/OL].武汉大学学报(信息科学版),1-8[2024-04-24].

[34] 冯明驰,高小倩,汪静姝,等.基于立体视觉与激光雷达的车辆目标外形位置融合算法研究[J].仪器仪表学报,2021,42(10):210-220.

[35] Wang H, Lou X, Cai Y, et al. Real-time vehicle detection algorithm based on vision and lidar point cloud fusion[J]. Journal of Sensors, 2019, 2019.

[36] 胡远志,刘俊生,何佳,等.基于激光雷达点云与图像融合的车辆目标检测方法[J].汽车安全与节能学报,2019,10(04):451-458.

[37] 李研芳,黄影平.基于激光雷达和相机融合的目标检测[J].电子测量技术,2021,44(05):112-117.

[38] Wang S. Wireless network indoor positioning method using nonmetric multidimensional scaling and RSSI in the Internet of things environment[J]. Mathematical Problems in Engineering, 2020, 2020: 1-7.

[39] 黄威,张玲华.基于RSSI的无线传感器网络室内定位算法研究[J].计算机技术与发展,2020,30(10):64-68.

[40] Oryad F E, Amindavar H. Wireless positioning based on hierarchical symbolic dynamic filtering of RSSI time series[J]. Signal Processing, 2023, 206: 108903.

[41] Shen X, Xu B, Shen H. Indoor Localization System Based on RSSI-APIT Algorithm[J]. Sensors, 2023, 23(24): 9620.

[42] Patwari N, Wilson J. Spatial models for human motion-induced signal strength variance on static links[J]. IEEE Transactions on Information Forensics and Security, 2011, 6(3): 791-802.

[43] Sun Z, Li S, Yang H. Precise disturbance localization of long distance fiber interferometer vibration senor based on an improved time–frequency variation feature extraction scheme[J]. Infrared Physics & Technology, 2024, 137: 105097.

[44] Iyengar S G, Varshney P K, Damarla T. On the detection of footsteps based on acoustic and seismic sensing[C]//2007 Conference Record of the Forty-First Asilomar Conference on Signals, Systems and Computers. IEEE, 2007: 2248-2252.

[45] Orr R J, Abowd G D. The smart floor: A mechanism for natural user identification and tracking[C]//CHI'00 extended abstracts on Human factors in computing systems. 2000: 275-276.

[46] 丁庆怀.基于分布式光纤传感技术的铁路沿线外部入侵监测系统[J].铁道通信信号,2019,55(09):66-68.

[47] 卢涛,邢军辉,王威.多传感器融合的入侵监控系统设计[J].单片机与嵌入式系统应用,2020,20(03):38-41.

[48] Li Y, Liu Y, Dong H, et al. Intrusion detection of railway clearance from infrared images using generative adversarial networks[J]. Journal of Intelligent & Fuzzy Systems, 2021, 40(3): 3931-3943.

[49] 张雷,鲍蓉,朱永红,等.基于CSA-ResNet的人员入侵检测方法[J].计算机工程,2023,49(04):297-302+311.

[50] Yang S, Bu F L. Design of frequency adjustable pulse in intelligent electronic fence system[C]//Proceedings of the 2nd International Conference on Telecommunications and Communication Engineering. 2018: 344-347.

[51] Tang Z J, Liu P L, Luo Q, et al. Main Control System Design for Balanced Tension-Type Electronic Fence[J]. Applied Mechanics and Materials, 2014, 687: 349-352.

[52] Tang Z, Li T, Leng X. The hybrid time-frequency detection method based on EEMD for balanced tension-type electronic fence[J]. Applied Technologies and Innovations, 2015, 11(2): 52-63.

[53] 迟金生,李小阳,张亚飞.视频电子围栏技术在黄骅港门机自动化作业中研究[J].电子技术与软件工程,2021,(21):75-77.

[54] 闫建星.煤矿井下图像去雾与综采工作面异常视觉检测方法研究[D].西安科技大学,2022.DOI:10.27397/d.cnki.gxaku.2022.000435.

[55] 吴春香,吴恋.面向视频的人员入侵检测方法研究及应用[J].计算机技术与发展,2022,32(04):116-121.

[56] 傅荟瑾,史天运,王瑞,等.基于改进YOLOv5的高铁周界入侵人员检测方法研究[J].铁道标准设计,2023,67(09):162-169.

[57] 朱金连,蒋海敏,杨玉捷,等.基于改进YOLOv3的海关监管场所人员入侵检测算法[J].无线互联科技,2022,19(02):122-126.

[58] 杨栋,黄文政,张秋亮,等.基于Faster-RCNN的站台端部人员入侵检测研究[J].铁路计算机应用,2020,29(02):6-11.

[59] 何昕谕,秦丽娟.基于结合改进体素滤波算法与高斯滤波算法研究[J].信息记录材料,2022,23(07):229-231.

[60] Zhao Q, Gao X, Li J, et al. Optimization algorithm for point cloud quality enhancement based on statistical filtering[J]. Journal of Sensors, 2021, 2021: 1-10.

[61] 唐佳林,林寿南,周壮,等.基于特征变换结合KD树改进ICP的快速点云配准方法[J].计算机科学,2023,50(S2):260-264.

[62] 徐勤功,郭杜杜.基于Retinex理论的暗光图像增强算法[J].中国科技论文,2023,18(11):1267-1274.

[63] 章联军,张鹏,陈芬,等.融合深度学习与多尺度Retinex的水下图像增强方法[J].光电子·激光,2023,34(08):823-832.

[64] Land E H, McCann J J. Lightness and retinex theory[J]. Josa, 1971, 61(1): 1-11.

[65] 刘晓阳,乔通,乔智.基于双边滤波和Retinex算法的矿井图像增强方法[J].工矿自动化,2017,43(02):49-54.

[66] Chang R, Zhang B, Zhu Q, et al. FFA-YOLOv7: Improved YOLOv7 Based on Feature Fusion and Attention Mechanism for Wearing Violation Detection in Substation Construction Safety[J]. Journal of Electrical and Computer Engineering, 2023, 2023.

[67] 张达为,刘绪崇,周维,等.基于改进YOLOv3的实时交通标志检测算法[J].计算机应用,2022,42(07):2219-2226.

[68] 宗伟林,陶洪钧,王琪,等.大功率掘进机危险区域报警系统设计与实现[J].煤炭科学技术,2014,42(09):72-76.

[69] 杨守瑞,段婉莹,艾文宇,等.光场相机建模与畸变校正改进方法[J].红外与激光工程,2023,52(01):239-247.