In the process of updating realistic 3D City, it is difficult to utilize background data. Image control points are usually dominated by image control targets and road markings, which vary greatly, leading to the risk of data reuse and difficulty in unifying the accuracy of two phases of results. A concept for constructing a feature image control library based on real scene 3D background data has been proposed. By utilizing real scene 3D background data, the positions and feature information of key points have been to extracted, and a feature image control library has been constructed. During the process of updating data for aerial triangulation, dense digital control points have been extracted from the feature image control library, and the accuracy and pass rate of aerial triangulation have been improved. By experimental verification and extraction of image and model feature points, a feature image control library has been constructed. It can effectively solve the inconsistent accuracy caused by changes in image control points, reduce the measurement and stitching time of image control points significantly, improve the automation level of aerial triangulation, and ensure the continuity of the real scene 3D updating process.