Localization of tagged objects with sufficient precision is a main issue in many industrial applications. In this paper, a new approach is proposed for the localization of UHF passive tags simply using the environment learning approach. This method uses, for a given passive tag located in a 2-D space, an aggregate function of the Received Signal Strength Indicator (RSSI) for all the possible RFID reader's transmission powers. Based on these measurements, we define a location signature, which is compared with those of other tags located at known positions in the neighborhood. We then implement a method based on the k-Nearest Neighbor (k-NN) algorithm to estimate the position of the target tag. Using a realistic test case involving seventy tags and four antennas, we show a very significant improvement of the localization accuracy in comparison with the results obtained using a single RSSI value.