In order to improve the performance of traditional fingerprint detection methods in shared bike tracking

inertial sensors are used to measure data. Particle filtering（ PF） method is a widely used sensor fusion algorithm. However

initialization and weighting precess are problematic in shared bicycle positioning system. A new PF scheme is proposed to generate steady and stable localization knowledge. A single hidden layer feedforward networks（ SLFNs） is used to simulate estimation and improvement of multiple probability PF performance

and to realize the discrimination of similar fingerprints. At the same time

the algorithm is initialized by random consistency sampling（ RANSAC） to reduce convergence time. Experimental results show that the tracking error of the proposed scheme is less than 1. 2 m

which is prior to the selected contrast method.