Devicefree localization (DFL) is to estimate the location of object without carrying any electronic device. In allusion to problems such as low reconstruction speed of traditional radio tomographic imaging (RTI), massive store capacity and time complexity of offline training in traditional fingerprint matching (FM) etc., this paper proposed a localization estimation method with non offline training based on the theory of Fresnel. Firstly change of received signal strength (RSS) of links under the Fresnel region will be calculated, then change of RSS will be modified for satisfying some reality constraints, reference fingerprint database will be established. By establishing fingerprint database, capacity of training fingerprint database can be reduced effectively. Error of actual environment will be taken into consideration and correction will be made to weaken the environmental disturbances. Finally Knearest neighborhood algorithm will be applied to the target positioning estimation. The simulation results show that this method not only can ensure the positioning speed while matching, but also can ensure positioning accuracy.