Abstract: The high-precision positioning technology of a rail-type patrol robot is an important research direction in the area of intelligent patrol inspection of belt conveyors. An excessively long mining belt conveyor and a complex working environment severely affect the positioning accuracy of patrol robots. This study aims to address the problems of poor adaptability to tracks and limited positioning accuracy of the positioning technology of rail-type patrol robots in the field of mining belt conveyor patrol inspection. Therefore, a high-precision positioning method based on a modified fusion of encoder and near field communication, abbreviated NFC, double sensors is proposed. This work analyzes the influence of track and track environment characteristics of the belt conveyor track patrol robot on the encoder coefficient. It also proposes a track segmentation principle based on the same characteristics of a track surface, providing a basis for the subsequent correction and fusion algorithm. A recursive positioning method of the absolute value encoder is constructed based on the data feedback characteristics carried by the robot. Through the historical positioning sensor data of robot operation, the encoder coefficients are modified according to sections and directions. Further, the encoder coefficient correction method based on recursive least squares is proposed to improve the adaptability of the encoder to the track. Hence, corresponding positioning methods are constructed according to the different positions of the robot’s track segments. At the end of the segment, the fusion positioning of the encoder and NFC data are realized based on the Kalman filtering algorithm to reduce the cumulative error of the encoder. In the segment, to improve the positioning accuracy of the encoder, the subsection and direction correction coefficient and real-time data of the encoder are used for recursive positioning. Therefore, combined with the positioning of each section of the track, the continuous high-precision positioning of the track-type patrol robot on the entire track can be realized. Moreover, an experimental platform is built for the proposed method to conduct physical testing. The modified fusion positioning method is compared with encoder positioning, RFID positioning, and fusion positioning based on encoder and NFC. The results of the correction experiment indicate that the modified fusion localization algorithm based on the encoder and NFC can effectively improve the adaptability of orbital inspection robot localization to the orbital environment. Meanwhile, the results of the modified fusion experiment indicate that the positioning method can improve the positioning accuracy of the orbital inspection robot. Therefore, the proposed positioning method can be applied to the application scenario of a long-distance mining belt conveyor patrol inspection.