煤炭仍是我国未来几年的主体能源，自动化综采工作面发展大幅度提升了采煤效率。巷道施工方面我国主要采用悬臂式掘进机配套单体锚杆钻机工艺，掘进效率低导致采掘失衡。但悬臂式掘进机结构紧凑、适应性好，以自动化、智能化提高其效率和成形截割质量已经成为行业共识。井下位姿精确测量技术是实现悬臂式掘进机智能化的首要条件。借助视觉测量技术、捷联掼导定位、机器人运动学等技术，多传感器信息融合解决煤矿井下采掘装备精确定位成为最佳方案。论文研究基于双目立体视觉与捷联惯导组合的掘进机位姿测量方法，构建视觉与惯导融合定位模型，利用扩展卡尔曼滤波技术结合松耦合策略实现悬臂式掘进机精确定位和姿态测量，对无人化掘进工作面发展具有重要意义。

针对煤矿井下高粉尘、低照度、背景复杂等因素严重影响成像稳定性问题，提出了基于双目立体视觉的悬臂式掘进机位姿测量方法，结合双目视觉立体校正技术对双目视图进行立体校正，利用区域生长原理和高斯拟合定位方法实现光斑区域的分割以及光斑中心定位，通过环形匹配策略实现前后两帧双目图像的精确匹配，最后基于三角测量原理获得相邻时域的两组3D空间坐标，通过3D-3D运动估计实现悬臂式掘进机的位姿估计。

针对双目视觉测量过程中可能出现图像失帧、目标遮挡等情况，影响视觉测量精度和稳定性的问题，提出基于扩展卡尔曼滤波技术的双目视觉与捷联惯导组合定位方法。通过研究捷联惯导测量原理及其位姿解算方法，并采用三子样的算法进行捷联惯导位姿求解；结合双目视觉测量悬臂式掘进机位姿机理，利用最小二乘法实现两者数据对准，采用扩展卡尔曼滤波技术实现两种定位数据融合，实现悬臂式掘进机机身位姿测量的准确度和可靠性。

最后，搭建试验平台验证上述理论和方法，评价系统测量性能。分别在静态与动态情况下，从测量距离、测量精度和稳定性三个方面进行实验验证。理论研究和实验结果表明，对于基于双目视觉的悬臂式掘进机位姿测量方法，随着测量距离的增大其测量性能有所下降，在实验所采用标靶的最大测量距离为7m，位置最大误差为73mm，满足煤矿井下定位需求。同时验证了动态情况下其位置和姿态的测量性能，对比发现其在X方向的测量误差为30mm，航向角误差在1°以内，满足井下要求。悬臂式掘进机双目视觉与捷联惯导组合定位方法使得位姿测量稳定性显著提升，很好的补偿了捷联惯导的累积误差问题，保证了自主连续性高精度测量。

视觉与惯导组合测量方法为实现井下移动装备精确定位提供了新的思路。鉴于悬臂式掘进机是煤矿井下运动最复杂的设备，其机身位姿精确测量技术对其它采掘装备具有良好借鉴作用，为智能化采掘奠定了技术基础。

In the future, coal is still the main energy of our country,and the  development of automatic fully mechanized mining face greatly improves the efficiency of coal mining. The technology of single bolt drilling rig for cantilever type roadheader is mainly used in the construction of roadway in China, which is the low efficiency way of roadheader leading to the imbalance of coal mining and excavation. But the cantilever type roadheader has compact structure and good adaptability, so it has become the consensus of the industry to improve the driving efficiency and ensure the forming quality by improving its automation and intelligence level. The precise measurement technique of position and posture is the first condition to realize the intelligent cantilever type roadheader. With the help of vision measurement technology, strapdown guide positioning, robot kinematics and other technologies, multi-sensor information fusion to solve the coal mine mining equipment accurate positioning has become the best scheme. Based on the combination of binocular stereo vision and strapdown inertial navigation, this paper constructs a fusion positioning model of vision and inertial navigation, and uses Extended Kalman Filter technology combined with loose coupling strategy to realize precise positioning and attitude measurement of cantilever roadheader, which is of great significance to the development of unmanned working face.

Aiming at the problems of high dust, low illumination and complex background which seriously affect the stability of imaging, a method for measuring the position of cantilever tunneling machine based on binocular stereo vision is proposed.Binocular Vision Stereo Calibration Using Binocular Vision Stereo Calibration；The segmentation of spot area and the location of spot center are realized by using region growth principle and Gaussian fitting method； The accurate matching of two frames binocular image is realized by ring matching strategy； Finally, two sets of 3D spatial coordinates in adjacent time domain are obtained based on the principle of triangulation. The pose estimation of cantilever roadheader is realized by 3D-3D motion estimation.

Aiming at the problems of image frame loss and target occlusion, which affect the accuracy and stability of visual measurement, this paper proposes a combination localization method based on extended Kalman filter. Establishing the error compensation of cone error and vortex effect model,and solving the agile inertial navigation pose using the three-sub-sample algorithm by studying the principle of strapdown inertial navigation measurement and its position and pose calculation method；Position and pose mechanism of cantilever roadheader measured by binocular vision, achieving the two data alignment with using the least square method, Fusing two localization data by extended Kalman filter to improve the accuracy and reliability of body position measurement of cantilever roadheader.

Finally, a test platform is set up to verify the above theory and method to evaluate the measurement performance of the system. The measurement distance, measurement accuracy and stability are tested respectively in the static and dynamic conditions, and considering the influence of the measurement of position and posture when the target moves simultaneously. Theoretical and experimental results indicate that when the measuring distance is 5 m, the position measurement error is less than 98 mm, the Z direction measurement error is less than 11 mm, the attitude angle measurement error is less than 1°, and the heading angle error is less than 0.2°and Improving the stability of position and pose measurement significantly, which compensates the cumulative error of strapdown inertial navigation and ensures the self-continuity and high precision measurement with the combined positioning method of binocular vision and strapdown inertial navigation of cantilever roadheader.

The combination measurement method of vision and inertial navigation provides a new idea for the accurate positioning of underground mobile equipment. In view of the fact that the cantilever tunneling machine is the most complicated equipment for underground movement in coal mine, the precise measurement technology of fuselage position and posture has a good reference for other mining equipment, which lays a technical foundation for intelligent mining.