Using GPS/INS data to enhance image matching for real-time aerial triangulation

• An image matching system that well suits real-time AT is presented.
• We present how GPS/INS data are used to enhance KLT algorithm.
• Error propagation is used to find number of pyramid levels for KLT tracking.
• We compare the accuracy and speed between the original KLT and the proposed work.
• The proposed system can improve the accuracy of EOs acquired from GPS/INS sensors.

Direct georeferencing is a promising technique for determining the exterior orientation parameters (EO) of a camera in real-time through the integration of GPS/INS sensors. Instead of using expensive devices, we improve the accuracy of the directly measured EOs through aerial triangulation (AT) and rely on tie-points. In this work, using GPS/INS data, we enhance the KLT tracker to achieve accuracy and speed that is compatible with real-time aerial triangulation. Given GPS/INS data from medium-grade sensors, the proposed system is 48% faster than the original work and tie-points extracted by our system are 6.33% more accurate and more evenly distributed than tie-points extracted by the original work. The AT processing results show that tie-points from the proposed work can reduce the RMSE of the directly measured EOs by 17.87% for position and 23.37% for attitude. Thus, we conclude that our proposed system can be integrated with real-time aerial triangulation.