Because RFID readings are inherently noisy due to the hardware limitations and the environmental interference, the applications of the RFID are restricted in many areas. In order to estimate the parameter of target symbol location, a Bayesian model was developed to inverse the parameter of target symbol location by taking use of observed RFID readings in multiple readers. The inference of the Bayesian model is based on 3-state detection model. A Markov Chain Monte Carlo (MCMC) method is used to sample the Bayesian joint posterior probability function to get the marginal posterior distributions of the target symbol location parameters, by compromising prior knowledge and constraints in RFID detection region. Simulated experiments have demonstrated the validity and accuracy of the method proposed to estimate symbol location.