Most of current wireless indoor positioning methods could not accurately obtain channel model, the mapping between spatial position and received signal features. The main factor for a precise channel model in an indoor environment is multipath effect. Time reversed (TR) wireless indoor positioning method has been validated to effectively reduce signals fading or time delay affected by multipath effect. However, these advantages are depended on a prior known channel model, without this condition, the accuracy of TR method will be seriously deteriorated. To solve the shortcoming of a general TR method in an unknown channel model application, we present a combining Time Reversal and Fast Marching Method (TR-FMM) positioning method. This method locates a target with two stages. In the stage one, the precise channel model of an indoor environment is estimated by FMM and simultaneous algebraic reconstruction technique (SART). In this stage, Time of Flight (TOF) information generated by some fixed spatial position anchors are used to fulfill the indoor channel model estimation, then the needed channel impulse response (CIR) for TR method will be obtained based on the estimated channel model. In the stage two, with the obtained CIR, any new joint mobile target will be accurately located by a general TR wireless indoor positioning method. Some numerical simulations have been presented to validate the proposed method. Simulative results depict the positioning deviation is less than 3cm for a newly joined mobile target with 1cm scale in a moderate complex indoor configure, and the accuracy of the positioning is improved 30 times comparing to a general TR method. The positioning time in the stage 2 is less than 3 minutes in a PC with 1.6 GHz dual CPUs and 2G Bytes memory. Obviously, the proposed method has great advantage in high accuracy and low complexity for wireless indoor positioning system.