掘锚自动化作业是煤矿巷道智能掘进的关键，针对当前掘锚设备交替作业过程中相对位姿测量和碰撞检测难题，提出一种数字孪生驱动的煤矿井下掘锚设备跟踪定位与碰撞检测方法。首先，为克服井下掘进工作面低照度、高粉尘、复杂背景干扰的影响，以多点红外LED标靶作为信息源，通过工业相机采集红外LED特征点图像，利用Hough轮廓检测与质心法提取光斑中心并通过二进制编码识别标靶ID，采用改进稀疏光流算法对光斑进行跟踪，同时建立基于PNP的掘锚设备位姿解算模型，采用对偶四元数获得设备间相对位姿。其次，利用数字孪生技术，基于Unity3D平台建立对应实际尺寸的掘锚设备及工作面数字孪生模型，利用Socket通信方式实现虚拟空间与物理实体之间的实时数据传输与交互，在虚拟空间中实现掘锚设备实时位姿的三维可视化，结合任意多边形OBB(oriented bounding box)碰撞检测算法，实现掘锚设备虚拟碰撞检测。最后，搭建实验平台进行掘锚设备位姿测量试验，同时对虚实运动轨迹和碰撞检测效果进行验证。实验结果表明：掘锚设备跟踪定位实验的位置误差不超过20 mm，角度误差不超过0.30°；虚实位置坐标对比中X轴方向最大误差不超过1.14 mm；Y轴方向最大误差不超过1.10 mm，能够保证系统虚实一致性和同步性，满足掘进工作面作业过程中掘锚设备实时跟踪定位及碰撞检测的要求。

The automation of excavating and anchoring operation is critical for intelligent excavation of coal mine roadway. Aiming at the challenges of relative pose measurement and collision detection during the alternating operation of the current excavating and anchoring equipment, a digital twin-driven tracking, positioning and collision detection method was proposed. Firstly, in order to overcome the influence of low illumination, high dust and complex background interference in the underground excavating face, the multi-point infrared LED target is taken as the information source, the infrared LED feature point image is collected by industrial camera, the center of the spot is extracted using the Hough contour detection and the center of mass method, the ID of the target is identified by binary coding, and the improved sparse optical flow method is adopted to track the spot. Meanwhile, a PNP-based pose solution model of excavating and anchoring equipment is established, and the relative pose of equipment is obtained using a dual quaternion. Secondly, the digital twin technology is utilized to establish the digital twin model of the excavating and anchoring equipment and working face at the actual size based on the Unity3D platform. The real-time data transmission and exchange between the virtual space and the physical entity is realized by Socket communication, the 3D visualization of the real-time pose of the excavating and anchoring equipment in the virtual space is achieved, and the virtual collision detection of the excavating and anchoring equipment is implemented in combination with the oriented bounding box (OBB) collision detection algorithm. Finally, an experimental platform is set up to complete the pose measurement test of the excavating and anchoring equipment, and at the same time, the virtual and real movement trajectories and collision detection effect are verified. The experimental results show that the position and angle errors of the tracking and positioning experiment of the excavating and anchoring equipment are less than 20 mm and 0.30° respectively. In the comparison of virtual and real position coordinates, the maximum error in the X-axis direction does not exceed 1.14 mm, the maximum error in the Y-axis direction does not exceed 1.10 mm, which can ensure the virtual and real consistency and synchronization of the system, meeting the requirements of real-time tracking, positioning and collision detection of the excavating and anchoring equipment during the operation of the excavating face