Passive target localization in electronic reconnaissance is often compromised by measurement noise, which significantly affects location accuracy. To address this challenge, this paper introduces a recursive least squares localization method that incorporates dynamic noise weighting. Initially, a two-dimensional motion model is formulated for both the target and the UAV platform. The observation equations are then constructed by integrating the angle of arrival (AOA) and the phase difference rate (DPR). Subsequently, building on error propagation theory, a time-varying noise variance model is developed, leading to the design of an adaptive noise covariance weighting matrix. Furthermore, a recursive weighted least squares estimation process is implemented through the iterative update of the Jacobian matrix and Kalman gain. The simulation results reveal that in a scenario where the target remains stationary and the UAV moves at a constant velocity, the proposed method achieves convergence within 250 s, with the final localization error within 100 m. Compared with the standard least square method, this result shows a significant performance improvement. By dynamically adjusting the observation weights via noise variance inversion, the method effectively mitigates the impact of interference during periods of high noise. Consequently, it offers a robust solution for target localization in complex electromagnetic environments.