Abstract: As the main equipment of fully mechanized mining face in coal mine, the degree of automation and intelligence of roadheader is of great significance to realize intelligent rapid excavation of roadway. Accurate and robust measurement method of the body posture of roadheader is the premise of intelligent control of roadheader. Underground dust and water mist, long and narrow confined space and unstructured complex environment have great influence on the accuracy and reliability of the posture measurement method of roadheader fuselage. In order to improve the accuracy and robustness of posture measurement of roadheader fuselage, a visual posture measurement method of roadheader based on infrared LED linear target is proposed. The infrared LED target with the characteristics of curve and straight line combination is designed, and the visual measurement system of roadheader position and posture is constructed. On the basis of analyzing the characteristics of the target image, an adaptive guided Hessian matrix target light bar center extraction method is proposed. The initial center and width of the light bar are obtained by image thinning and Gaussian fitting, the sub-pixel coordinates of the target light bar center are obtained by Hessian matrix, and the ellipse and straight line equations are obtained by least square fitting. A monocular vision pose calculation model of roadheader fuselage based on the combination of curve and straight line is constructed, and the optimal pose estimation is realized by minimizing the projection error, and the pose information of roadheader fuselage is obtained by coordinate system transformation. The influence of different factors on positioning accuracy is analyzed by numerical simulation, and the optimal target size and error distribution at different distances are determined. An experimental verification platform was built, and the maximum measurement errors of X-axis, Y-axis and Z-axis were 58.820, 69.525, 133.2 mm respectively within the 40-meter measurement range. Experiments show that this method effectively overcomes the influence of low illumination, stray light background and occlusion, and has an important reference for the spatial long-distance pose detection of underground tunnels in coal mines.