The invention discloses a compressed sensing mine target positioning method based on discrete dynamic mesh selection for mainly the problem of reducing the influence of electromagnetic interference onthe positioning accuracy in a mine environment, so that the positioning accuracy is significantly improved. The method comprises the following steps: in an offline state, sampling signal strength indiscrete regions to obtain a feature fingerprint database; and transmitting signals in an online state, performing signal strength sampling on the target at a reference location to obtain a measurement value vector, in a first positioning phase, constructing a measurement matrix via a reference node location, importing the measurement matrix into the fingerprint database from a server, constructing a sparse base, and performing dynamic selection on divided grid areas via a reconstruction algorithm under the condition that the measurement matrix and the sparse base met irrelevant conditions toobtain several optimal grid locations, in a second positioning phase, obtaining corresponding coordinates of the several grid areas obtained based on compressed sensing positioning via the fingerprintdatabase, selecting a corresponding weighting function, and accurately calculating a final location of the target by using a grid centroid method.