针对煤矿巷道复杂环境下掘进机精确定位难题，提出了一种惯导与视觉融合的掘进机精确定位方法，该方法采用惯导与“视觉+激光标靶”的定位方案。该方案将设计的四特征点大尺寸激光标靶固定于巷道顶板，相机固定于掘进机机身采集激光标靶图像，并运用圆拟合法定位光斑中心和基于四特征点的EPnP算法解算掘进机位置。为了验证“视觉+激光标靶”方法对掘进机位置检测效果，在模拟掘进工作面环境下开展了“视觉+激光标靶”位置检测试验，结果表明：在30 m内沿巷道宽度方向、掘进方向、高度方向最大误差不超过28.549 mm、78.868 mm、44.459 mm，实现了掘进机位置精确检测。针对惯导测量掘进机位姿误差随时间累积和掘进机振动对组合定位系统产生干扰导致位姿检测不准问题，提出改进Sage-Husa自适应滤波的惯导与视觉信息融合方法，该方法通过检测新息方差值修正量测误差来提高定位准确性。在模拟掘进工作面环境下开展了惯导与“视觉+激光标靶”组合定位实验，采用改进前后Sage-Husa自适应滤波算法融合惯导与视觉信息进行对比分析，结果表明：改进后Sage-Husa自适应滤波算法融合得到的定位误差更小，俯仰角、横滚角、航向角最大误差分别为0.029°、0.051°、0.011 3°，在距离激光标靶30 m内巷道宽度位置误差在0.033 m范围内，巷道掘进方向位置误差在0.062 m范围内。所提出的惯导与视觉融合定位方法能够满足巷道掘进定位精度要求。

For the difficulty of pinpointing of tunneling machines in the complex environment of coal mine tunnels,a precise positioning method of roadheader based on inertial navigation and visual fusion is proposed, which uses a positioning scheme of inertial navigation plus "vision+laser target". With the designed four feature point large-sized laser target fixed on the roof of the tunnel and the camera on the tunneling machine to collect laser target images, the scheme uses the circular fitting method to locate the center of the light spot and the EPnP algorithm based on four feature points to calculate the position of the tunneling machine. A “vision+laser target” position detection experiment was conducted in simulated condition of heading face to verify the effectiveness of the method in detecting the position of tunneling machines. The results are as follows: within 30 meters tunnel, the maximum error in the width, position, and height are no more than 28.549, 78.868, and 44.459 mm, which shows accurate detection of the position of the tunneling machine. For the inaccurate pose detection of tunneling machines in inertial navigation measurement, induced by the accumulation of pose errors over time and the machine vibration on the combined positioning system, an improved Sage-Husa adaptive filtering method for inertial navigation and visual information fusion is proposed, which could correct measurement errors by detecting new variance values to improve positioning accuracy. A combined positioning experiment of inertial navigation plus “vision+laser target” was conducted in a simulated heading face environment, using improved Sage-Husa adaptive filtering algorithm to fuse inertial navigation and visual information for comparative analysis, which show that the improved Sage-Husa adaptive filtering algorithm fusion results in smaller positioning errors. The maximum errors of pitch angle, roll angle, and heading angle are 0.029°, 0.051°, and 0.0113°, respectively. Within 30 meters, the position error of tunnel width is within the range of 0.033 m, and the position error of tunnel excavation direction is within the range of 0.062 m. Overall, the proposed inertial navigation plus visual fusion positioning method can meet the accuracy requirements of tunnel excavation positioning.