Visual Feature Extraction And Simultaneous Localization And Mapping Of Unmanned Aerial Vehicle

In recent years, UAV(Unmanned Aerial Vehicle) has been greatly improved with the improvement of the science and technology, and has played an increasingly important role in military and civilian. In the future, UAV will be fully autonomous flight, and autonomous navigation is one of the most important technologies. Traditionally, the UAV's position and attitude information can be obtained from Inertial Navigation System(INS) and Global Positioning System(GPS). However, GPS will not work when signals are too weak, and INS will stop working when it causes a large accumulated errors. Image sensors are being widely used in the field of navigation due to their lower weight, and can get a great deal of information. So in this article, SLAM autonomous navigation algorithm for UAV was mainly studied, the main works are given as following:First, the difference between the two algorithms of SIFT and SURF is analyzed theoretically. Using integral image and Hessian in feature extraction, SURF has less running time, and this conclusion was proved by the experiment.Secondly, the EKF-SLAM algorithm was studied which is based on extended Kalman filter algorithm. Simultaneous localization and mapping can be realized with the algorithm at some accuracy, but it is not suitable to noisier macro environment because its complexity and bad robustness, which was found by the simulations.Then, the FastSLAM algorithm was introduced to improve the accuracy and robustness of the EKF-SLAM. This algorithm estimates the UAV flight paths with particle filters and estimates the feature points for each particle by EKF algorithm. With the advantages of the particle filter, the robust stability and accuracy of the algorithm are improved.Finally, an adaptive resampling algorithm was proposed to solve the problem of bad real-time performance due to excessive resampling in the SLAM algorithm. And two kinds of heuristic factors are introduced as the guiding factor of particle swarm optimization algorithm. The diversity of the particle swarm and the accuracy of the algorithm were improved by optimizing the particles with the heuristic factors after resampling. The simulation experiments show that the algorithm in this paper has been improved in accuracy and robustness.