Cross-view image geo-localization aims to determine the geographical location by matching ground images with satellite images that have geographic tags. Existing deep learning-based methods use convolutional neural networks (CNNs) or Vision Transformers to extract features from both ground and satellite images, representing these features with high-dimensional floating-point numbers. This leads to high computational costs and memory usage when calculating feature similarities, making it difficult to meet real-time requirements. To address this issue, this paper proposes a two-stage network model. In the first stage, a feature extraction network generates high-dimensional floating-point features with strong representation capability; in the second stage, a novel hash encoder is designed to effectively map high-dimensional features to compact binary hash codes, significantly reducing computational complexity and memory usage. Experimental results show that the proposed method performs excellently on the CVUSA and CVACT datasets, not only improving matching efficiency but also maintaining high localization accuracy. This study provides an effective solution for the real-time application of cross-view image geo-localization.