In order to achieve high-precision building information extraction, a rule-based object-oriented method is adopted, and a technical process of image preprocessing, multi-scale segmentation, construction of rule information, and feature extraction is proposed. Two sample areas in Yangzhou City, China (Jiajia Garden Community and Lianyi Nanyuan Community) based on the domestically produced GF-2 satellite are used for experimental validation of the method. It is shown that compared with the traditional method, the new method has better extraction effect and higher accuracy, with a recognition accuracy of 97.7% and a Kappa coefficient of 0.93.