<p>近年来，煤矿智能化已经成为煤炭行业高质量发展的核心理念，建设智能化掘进工作面和提高掘进装备的位姿检测技术已经发展为实现智能化建设的关键技术之一。计算机技术和机器视觉技术的快速发展，使得利用视觉感知和智能算法实现动态目标的非接触位姿测量已经成为煤炭领域研究热点。因此，针对煤矿井下复杂、多变环境中掘进装备视觉位姿测量技术，论文提出了一种基于点线特征的视觉跟踪与定位方法，研究了掘进工作面视觉信息图像增强方法、视觉点线特征区域跟踪方法和掘进装备位姿解算模型等技术，构建了针对巷道掘进装备的视觉点线特征定位系统，实现了掘进装备巷道空间全局位姿测量。论文主要研究工作包括：</p> <p>针对掘进工作面由于光照较低、亮度不均和噪声较多等因素导致图像视觉信息显著性差问题，提出了一种掘进工作面视觉信息图像增强方法。研究可分离高斯滤波进行原始图像一维水平与垂直方向卷积降噪，利用完美反射法和混合亮度实现图像全局自动白平衡与局部细节增强。使用递归区域分割方法将图像划分为高光区、中光区和低光区，并采用改进的大气散射模型进行图像快速去雾。通过拉普拉斯锐化操作与加权混合滤波处理，增加图像高频成分，抑制图像低频成分，提高图像整体对比度，抑制图像轮廓或边缘灰度发生突变，为掘进装备视觉跟踪与定位提供高质量的信息源。</p> <p>针对掘进装备视觉定位方法中定位特征标定困难、杂光干扰强和特征提取稳定性差等问题，提出了一种基于视觉注意力机制的特征区域跟踪方法。以矿方激光指向仪作为全局坐标引入基准进行视觉定位特征标定，利用防爆工业相机和镜头参数建模分析，确立掘进装备视觉定位特征区域，进行初始帧图像区域特征提取与融合研究，建立特征区域模型，并引入视觉注意力机制对后续帧图像进行自适应候选区域框生成。根据特征区域模型与自适应区域候选框之间的欧式距离建立候选区域图模型，利用Battacharyya系数基于候选区域图模型进行特征区域与候选区域相似性度量，实现视觉信息图像特征区域的跟踪，提高基于特征跟踪区域进行特征提取的稳定性和准确性。</p> <p>针对掘进装备视觉定位时相机视野易受遮挡和定位稳定性差等问题，提出了一种基于跟踪区域点线特征的掘进装备位姿解算方法。通过对两个单目防爆工业相机进行视场角和空间布局建模分析，构建一种基于点线特征的掘进装备视觉位姿测量系统。研究改进RANSAC算法进行特征跟踪区域点特征提取与线特征拟合，获得跟踪区域特征标靶点线定位特征，并利用N型多点三线掘进装备位姿解算模型进行掘进装备机身位姿求解，获得掘进装备相对于特征标靶坐标系的空间位姿信息。研究掘进工作面掘进装备空间位姿转换，实现掘进装备巷道空间全局机身位姿测量，提高掘进装备视觉位姿测量系统的稳定性与鲁棒性。</p> <p>最后，在上述理论和方法研究的基础上，开发了掘进装备视觉跟踪与定位系统软件平台。在狭长空间环境下，模拟煤矿井下环境，搭建实验平台，对系统的各功能模块进行功能验证与性能测试。实验结果表明，掘进装备视觉跟踪与定位系统在狭长空间环境下的位置误差为&plusmn;30mm，姿态误差为&plusmn;1.0&deg;，有效保证了煤矿井下复杂、多变环境中掘进装备视觉位姿测量的鲁棒性，对促进智能化掘进技术发展具有重大意义。</p>

<p>In recent years, coal mine intelligence has become the core concept of high-quality development of the coal industry. The construction of intelligent heading face and the improvement of pose detection technology of heading equipment have developed into one of the key technologies to realize intelligent construction. With the rapid development of computer technology and machine vision technology, non-contact pose measurement using visual perception and intelligent algorithms to achieve dynamic targets has become a research hotspot in the coal field. Therefore, in view of the visual pose measurement technology of tunneling equipment in complex and changeable environment of underground coal mine, this thesis proposes a visual tracking and positioning method based on point and line features, studies the visual information image enhancement method of tunneling face, the visual point and line feature region tracking method and the pose calculation model of tunneling equipment, constructs the visual point and line feature positioning system for tunneling equipment, and realizes the global pose measurement of tunneling equipment roadway space. The main research work includes :</p> <p>Aiming at the problem of poor saliency of image visual information caused by low illumination, uneven brightness and more noise in tunneling face, an image enhancement method of visual information in tunneling face is proposed. The separable Gaussian filter is used to perform one-dimensional horizontal and vertical convolution noise reduction of the original image. The perfect reflection method and mixed brightness are used to achieve global automatic white balance and local detail enhancement of the image. The image is divided into high light area, medium light area and low light area by recursive region segmentation method, and the improved atmospheric scattering model is used to quickly remove the fog. Through Laplace sharpening operation and weighted hybrid filtering processing, the high frequency component of the image is increased, the low frequency component of the image is suppressed, the overall contrast of the image is improved, and the mutation of the image contour or edge gray level is suppressed, which provides high-quality information source for visual tracking and positioning of tunneling equipment.</p> <p>Aiming at the problems of difficult calibration of positioning features, strong stray light interference and poor stability of feature extraction in visual positioning method of tunneling equipment, a feature region tracking method based on visual attention mechanism is proposed. The miner &#39;s laser pointer is used as the global coordinate introduction benchmark to calibrate the visual positioning feature. The explosion-proof industrial camera and lens parameter modeling analysis are used to establish the visual positioning feature area of the tunneling equipment. The feature extraction and fusion of the initial frame image region are studied, and the feature region model is established. The visual attention mechanism is introduced to generate the adaptive candidate region box for the subsequent frame image. The candidate region graph model is established according to the Euclidean distance between the feature region model and the adaptive region candidate box. The Battacharyya coefficient is used to measure the similarity between the feature region and the candidate region based on the candidate region graph model, so as to realize the tracking of the feature region of the visual information image and improve the stability and accuracy of feature extraction based on the feature tracking region.</p> <p>Aiming at the problems of easy occlusion of camera vision and poor positioning stability in visual positioning of tunneling equipment, a pose calculation method of tunneling equipment based on point and line features of tracking area is proposed. By modeling and analyzing the field of view and spatial layout of two monocular explosion-proof industrial cameras, a visual pose measurement system for tunneling equipment based on point-line features is constructed. The improved RANSAC algorithm is used to extract the point feature and fit the line feature in the feature tracking area, and the point and line positioning features of the feature target in the tracking area are obtained. The pose calculation model of the N-type multi-point three-line tunneling equipment is used to solve the pose of the tunneling equipment, and the spatial pose information of the tunneling equipment relative to the feature target coordinate system is obtained. The spatial pose transformation of tunneling equipment in tunneling face is studied to realize the global pose measurement of tunneling equipment roadway space and improve the stability and robustness of visual pose measurement system of tunneling equipment.</p> <p>Finally, on the basis of the above theory and method research, the software platform of visual tracking and positioning system for tunneling equipment is developed. In the narrow space environment, the experimental platform is built to simulate the underground environment of coal mine, and the functional verification and performance test of each functional module of the system are carried out. The experimental results show that the position error of the visual tracking and positioning system of the tunneling equipment is &plusmn;30 mm and the attitude error is &plusmn;1.0 &deg; in the narrow and long space environment, which effectively ensures the robustness of the visual pose measurement of the tunneling equipment in the complex and changeable environment of the coal mine. It is of great significance to promote the development of intelligent tunneling technology.</p>

[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] 杨健健, 张强, 王超, 等. 煤矿掘进机的机器人化研究现状与发展[J]. 煤炭学报, 2020, 45(08): 2995-3005.

[4] 陶云飞, 杨健健, 李嘉赓, 等. 基于惯性导航技术的掘进机位姿测量系统研究[J]. 煤炭技术, 2017, 36(01): 235-237.

[5] Schneider O. Requirements for positioning and navigation in underground constructions [C]// 2010 International Conference on Indoor Positioning and Indoor Navigation. IEEE, 2010: 1-4.

[6] 陶云飞, 李瑞, 李嘉赓, 等. iGPS的单站多点分时测量系统对掘进装备偏向位移精度研究[J]. 煤炭技术, 2017, 36(02): 246-247.

[7] 马源, 符世琛, 张子悦, 等. 悬臂式掘进装备位姿检测方法研究现状[J]. 工矿自动化, 2020, 46(08): 15-20.

[8] 李睿. 悬臂掘进机机身位姿参数检测系统研究[D]. 北京: 中国矿业大学, 2012.

[9] 田原. 悬臂掘进机自动导向和定位技术探索[J]. 工矿自动化, 2010, 36(08): 26-29.

[10] Fu S, Li Y, Zhang M, et al. Ultra-wideband pose detection system for boom-type roadheader based on Caffery transform and Taylor series expansion[J]. Measurement Science and Technology, 2017, 29(01): 015101.

[11] 杨文娟, 张旭辉, 张超, 等. 基于三激光束标靶的煤矿井下长距离视觉定位方法[J]. 煤炭学报, 2022, 47(02): 986-1001.

[12] 胡兴涛, 朱涛, 苏继敏, 等. 煤矿巷道智能化掘进感知关键技术[J]. 煤炭学报, 2021, 46(07): 2123-2135．

[13] 刘晓阳, 刘毅. 改进的脉冲耦合神经网络矿工图像增强方法[J]. 煤炭学报, 2011, 36(S1): 207-210.

[14] 王洪栋, 郭伟东, 朱美强, 等. 一种煤矿井下低照度图像增强算法[J]. 工矿自动化, 2019, 45(11): 81-85.

[15] 蔡利梅, 钱建生, 赵杰, 等. 基于模糊理论的煤矿井下图像增强算法[J]. 煤炭科学技术, 2009, 37(08): 94-96.

[16] 智宁, 毛善君, 李梅. 基于照度调整的矿井非均匀照度视频图像增强算法[J]. 煤炭学报, 2017, 42(08): 2190-2197.

[17] 樊占文, 刘波. 基于改进的Retinex低照度图像自适应增强技术研究[J]. 工矿自动化, 2021, 47(S1): 126-130.

[18] 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.

[19] Dai L, Qi P, Lu H, et al. Image enhancement method in underground coal mines based on an improved particle swarm optimization algorithm[J]. Applied Sciences, 2023, 13(05): 3254.

[20] 付燕, 李瑶, 严斌斌. 一种煤矿井下视频图像增强算法[J]. 工矿自动化, 2018, 44(07): 80-83.

[21] 尚长春, 马宏伟, 张奇. 煤矿井下机器视觉图像增强技术研究[J]. 计算机测量与控制, 2013, 21(10): 2777-2779.

[22] 洪炎, 朱丹萍, 龚平顺. 基于TopHat加权引导滤波的Retinex矿井图像增强算法[J]. 工矿自动化, 2022, 48(08): 43-49.

[23] 张英俊, 雷耀花, 潘理虎. 基于暗原色先验的煤矿井下图像增强技术[J]. 工矿自动化, 2015, 41(03): 80-83.

[24] 张立亚, 郝博南, 孟庆勇, 等. 基于HSV空间改进融合Retinex算法的井下图像增强方法[J]. 煤炭学报, 2020, 45(S1): 532-540.

[25] 王媛彬, 韦思雄, 段誉, 等. 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.

[26] 田子建, 王满利, 吴君, 等. 基于双域分解的矿井下图像增强算法[J]. 光子学报, 2019, 48(05): 107-119.

[27] 杨志龙, 侯志强, 余旺盛, 等. 一种时空特征融合的鲁棒视觉跟踪算法[J]. 空军工程大学学报, 2022, 23(06): 57-63.

[28] 余旺盛, 田孝华, 侯志强, 等. 基于关键区域特征匹配的视觉跟踪算法[J]. 电子学报, 2014, 42(11): 2150-2156.

[29] 任世杰, 杨小冈, 齐乃新, 等. 一种基于局部特征索引结构的目标跟踪方法[J]. 电光与控制, 2019, 26(08): 84-89.

[30] Zinßer T, Gräßl C, Niemann H. Efficient feature tracking for long video sequences[C]// Pattern Recognition: 26th DAGM Symposium, Tübingen, Germany, August 30-September 1, 2004. Proceedings 26. Springer Berlin Heidelberg, 2004: 326-333.

[31] 李中科, 万长胜. 具备重检测机制的融合特征视觉跟踪算法[J]. 图学学报, 2018, 39(05): 892-900.

[32] 吴世东, 鲍华, 张陈斌, 等. 基于视觉显著性特征的粒子滤波跟踪算法[J]. 中国科学技术大学学报, 2015, 45(11): 934-942.

[33] 卞志国, 金立左, 费树岷. 特征融合与视觉目标跟踪[J]. 计算机应用研究, 2010, 27(04): 1537-1543.

[34] 陈建伟, 贺继林. 一种改进的多通道特征视觉跟踪算法[J]. 传感器与微系统, 2020, 39(03): 120-123.

[35] 郭利, 周盛宗. 基于通道权重融合特征的目标跟踪算法[J]. 计算机系统应用, 2020, 29(12): 178-186.

[36] Yu R, Fang X, Hu C, et al. Research on Positioning Method of Coal Mine Mining Equipment Based on Monocular Vision[J]. Energies, 2022, 15(21): 8068.

[37] 张旭辉, 沈奇峰, 杨文娟, 等. 基于三激光点标靶的掘进机机身视觉定位技术研究[J]. 电子测量与仪器学报, 2022, 36(06): 178-186.

[38] 杜雨馨, 刘停, 童敏明, 等. 基于机器视觉的悬臂式掘进机机身位姿检测系统[J]. 煤炭学报, 2016, 41(11): 2897-2906.

[39] 程健. 煤矿巷道机器人管线视觉辅助定位与导航方法研究[J]. 煤炭科学技术, 2020, 48(07): 226-232.

[40] 田原. 悬臂式掘进机视觉定位方法研究[J]. 矿山机械, 2019, 47(03): 8-12.

[41] 王学, 周红旭, 张雷, 等. 基于近红外双目立体视觉的悬臂式掘进机定位研究[J]. 工矿自动化, 2022, 48(07): 43-57.

[42] 薛旭升, 张旭辉, 毛清华, 等. 基于双目视觉的掘进机器人定位定向方法研究[J]. 西安科技大学学报, 2020, 40(05): 781-789.

[43] 张超. 悬臂式掘进机双目视觉与捷联惯导组合定位技术研究[D]. 西安: 西安科技大学, 2020.

[44] Wan J, Zhang X, Zhang C, et al. Vision and Inertial Navigation Combined-Based Pose Measurement Method of Cantilever Roadheader[J]. Sustainability, 2023, 15(05): 4018.

[45] 黄东, 杨凌辉, 罗文, 等. 基于视觉/惯导的掘进机实时位姿测量方法研究[J]. 激光技术, 2017, 41(01): 19-23.

[46] 刘传军, 韩冰, 杨洪民. 我国悬臂式重型掘进机的现状与发展趋势[J]. 煤炭技术, 2009, 28(11): 1-2.

[47] 刘仲民, 李战明. 基于模糊神经网络的双目视觉定位系统研究[J]. 计算机工程与设计, 2010, 31(03): 599-601.

[48] 丁震. 我国煤巷掘进装备现状及发展趋势[J]. 工矿自动化, 2014, 40(04): 23-27.

[49] 杜雨馨. 矿井悬臂式掘进机位姿感知及定位方法研究[D]. 北京: 中国矿业大学, 2019.

[50] Chen Z, Wang Z, Sheng B, et al. Dynamic RGB‐to‐CMYK conversion using visual contrast optimisation[J]. IET Image Processing, 2017, 11(07): 539-549.

[51] Nayar S K, Narasimhan S G. Vision in bad weather[C]//Proceedings of the seventh IEEE international conference on computer vision. IEEE, 1999, 2: 820-827.

[52] He K, Sun J, Tang X. Single image haze removal using dark channel pROIr[J]. IEEE transactions on pattern analysis and machine intelligence, 2010, 33(12): 2341-2353.

[53] Dong X, Pang Y, Wen J. Fast efficient algorithm for enhancement of low lighting video[M]// ACM SIGGRAPH 2010 Posters. 2010: 1-1.

[54] 周传德, 王见, 尹爱军. 基于平面标靶的物体空间姿态精确测量[J]. 重庆大学学报, 2011, 34(08): 66-70.

[55] 张旭, 郑泽龙. 主动光立体标靶检测及其在工业机器人位姿测量中的应用[J]. 中国机械工程, 2016, 27(19): 2594-2601.

[56] 黄喆,李佳雄,杜文阳,等.基于双屏视觉标靶的掘锚机机身位姿测量技术研究[J/OL].煤炭学报, 2022-07-07:

[57] 吕耀宇. 空间合作目标单目视觉位姿测量技术研究[D]. 长春: 中国科学院大学, 2018.

[58] 汤帆扬. 实时人脸检测及对齐技术的研究[D]. 杭州: 浙江工业大学, 2017.

[59] 张博强. 融合视觉注意力机制的相关滤波目标跟踪研究[D]. 阜新: 辽宁工程技术大学, 2022.

[60] Bhattacharyya A. On a measure of divergence between two statistical populations defined by their probability distribution[J]. Bulletin of the Calcutta Mathematical Society, 1943, 35: 99-110.

[61] Derpanis K G. Overview of the RANSAC Algorithm[J]. Image Rochester NY, 2010, 4(01): 2-3.

[62] 叶文龙. 基于多相机系统的机器人定位[D]. 杭州: 浙江大学, 2020.