Abstract: Mine target positioning has basically completed the transition from RFID technology to TOA or RSS ranging three-point solution method, but the three-point solution positioning method of electromagnetic wave ranging is faced with difficult problems that are difficult to be overcome in hardware systems and sensitive to electromagnetic noise in mine roadway.However, combining the theory of compressed sensing with localization can solve this problem very well.In this paper, a method of compressed sensing positioning based on discrete dynamic grid and the fingerprint feature library of signal intensity is proposed.Firstly, the positioning area is discretely meshed, and the signal strength characteristics at each grid position are measured offline to construct a feature fingerprint database.Then, the L-nearest neighbor dynamic node selection method is used to determine the target positioning to achieve regional positioning.Finally, the measurement matrix is constructed based on the fingerprint dictionary and the results of regional positioning, and the signal intensity field matrix is reconstructed by the compressed sensing algorithm to achieve accurate target location.The experimental results show that the method has a good positioning effect in the roadway environment with a higher positioning accuracy and less communication overhead.