Research On The Methods Of Landmark Correction For INS Based On Dynamic Vision Localization

Inertial navigation system (INS) is widely used because it can work independently. Because the location error drift goes unbounded with time, the INS can't work independently in high accuracy for a long time. Integrated navigation system is used to solve this problem, in which INS/Vision integrated navigation is widely used.The sequential images taken by camera in real-time are used to calculate motion parameters of moving vehicle to provide the navigation information in Vision navigation system. Vision navigation technology is now sucessfully used in robot navigation, autonomous vehicles, unmanned helicopters and so on.In the INS/Vision integrated navigation system based on the landmark, Vision system is used to correct the error of INS and prevent error drift of the INS goes unbounded, in this thesis. So vision system should execute three steps: first, identify landmark; second, dynamically calculate the location of INS based on landmark; third, correct the error of INS using Kalman filter.In the process of landmark identification, the image of landmark in camera compared with template image presented rotation transform and scale transform. The recognition algorithm based on log-polar transform was discussed. According to the actual application, the rotation transform was resolved by using the output of INS. And two methods were proposed to resolve scale problem, experiment results showed that this algorithm was feasible to meet the actual needs.Dynamic localization algorithm of vision system has high demands on efficiency, because it's a short time for moving vehicle to pass the landmark. Traditional localization methods of triangle localization and trilateral localization can't meet this demand. A fast localization algorithm based on single landmark was presented, which can locate the vision system using only one frame image by the assistance of INS within less time. In the end, the output delay of INS and the vision system's acquisition and transmission delay were measured by experimental calibration, solved the integrated navigation data synchronization of the INS and the vision system.The error model of INS was also analyzed in the thesis, so that a seven ststes Kalman filter was constructed. Considering the long period of landmark correction, asynchronous Kalman filter is used to correct the error of INS. Simulation results showed that the Kalman filter algorithm had good effect on correcting the error of INS.