In order to construct a radio environmental map that can help navigating secondary communication in TV bands, we present a new method of interference source localization for cognitive radio networks. Our low-cost, high-precision localization method, operates without any prior knowledge of the interference source other than transmitter power. It can be divided into two steps: an estimation step at each sensor based on classical methods to estimate the angle of arrival and the received signal power, followed by an SNR-aided fusion. Simulation results show that tens of sensor nodes each having two antennas in a large-scale area setting are enough to achieve accurate location. The proposed location method is practical, economic and essential for blind interference source localization in cognitive radio networks.