煤矿井下掘进装备长距离动态定位是巷道掘进智能化发展面临的首要难题，准确、可靠的位姿测量与估计对提高掘进效率具有重要意义。由于煤矿环境中存在振动噪声、磁干扰、静电干扰等因素，传统测量方法的稳定性和准确性难以保证。与其他定位方法相比，视觉测量方法以其无接触、无累积误差的优点备受关注。针对井下视觉定位面临的精度和稳定性难题，提出基于多激光束的煤矿井下长距离视觉定位方法，采用矿用激光指向仪构建了三激光束标靶解决低照度、高粉尘和复杂背景的干扰问题。以悬臂式掘进机为应用对象，构建了基于三激光束标靶的掘进机机身位姿单目视觉测量系统，提出基于颜色分量的峰值聚类约束的激光束标靶图像分割方法，利用Hessian矩阵求解激光束粗略中心线像素点的法线方向，结合泰勒展开实现三激光束中心线的亚像素级特征提取和定位；构建了三点三线（3P3L）掘进机机身位姿测量模型，并建立了测量模型的最小二乘法损失函数，结合最小化重投影误差实现非迭代全局最优解估计，获得了掘进机机身位姿的最优解，并对测量模型主要参数的误差分布进行了数值仿真研究。最后，搭建基于三激光束标靶的掘进机机身定位系统平台并进行实验评估，结果表明，该方法可满足煤矿井下恶劣环境下的精确定位需求。

Long distance dynamic positioning of coal mine heading equipment has become the primary problem for an intelligent development of roadway. Accurate and reliable pose measurement and estimation is of great significance to improve heading efficiency. It is difficult to guarantee the stability and accuracy of the traditional measurement method because of the vibration and noise, magnetic and electrostatic interference in the coal mine environment. Compared with other positioning methods, visual measurement method has attracted much attention for its advantages of no contact and no cumulative error. In this paper, aiming at the problems of accuracy and stability of underground visual positioning, a long distance visual positioning method based on multiple laser beams is proposed in coal mine, and a three laser beams target is constructed by mining laser point instrument to solve the interference problems of low illumination, high dust and complex background. The boom type roadheader is taken as the application object, and the monocular vision measurement system of the roadheader body is constructed based on a triple laser beams target. The method of laser beam target image segmentation by color component and peak value clustering constrained is proposed. The Hessian matrix is used to solve the normal direction for each pixel points of the rough center line of the laser beam. The sub pixel feature extraction and location of the center line of the three laser beams are realized by Taylor expansion. A three point three line (3P3L) roadheader body pose estimation model is constructed, the least square loss function of the measurement model is established, the non iterative global optimal solution is estimated by minimizing the re projection error, and the optimal solution of the body pose of the roadheader is obtained. Furthermore, the error distribution of the main parameters of the measurement model is studied by numerical simulation. Finally, the roadheader body positioning system platform based on the triple laser beams target is built and evaluated experimentally. The results show that this method can meet the requirement of precise positioning in the harsh environment of coal mine.

1 掘进设备位姿视觉测量系统

2 掘进机机身三点三线位姿解算模型

3 激光束标靶图像分割与特征提取

   3.1 三激光束标靶图像特性

   3.2 三激光束标靶图像分割与特征提取

4 掘进机机身位姿视觉测量精度分析

5 掘进机视觉测量系统性能测试

   5.1 测试系统的环境适应性实验

   5.2 机身位姿视觉测量系统性能实验

6 结论