Lunar Addressing And Navigation Based On Computer Vision

The Lunar exploration program is closely related to Chinese economic and military interests.Simulated lunar exploration project as the pre-research project of lunar exploration program has always been the hot spot in the academic research area,including lunar navigation,soft landing and so on.This paper mainly introduces visual navigation of lunar navigation.In order to improve the accuracy and stability of navigation,visual navigation can be divided into large area navigation scenarios and accurate navigation in ground fine scene.The main content is divided into the following two blocks.Under the big scene,this paper introduces the density based landmark selection navigation algorithm and the visual saliency based on landmark selection navigation.In the first navigation scheme,this paper designs a parameter adaptive DBSCAN clustering algorithm.The algorithm time complexity is only NLogN which means it is real-time.In the second navigation scheme,saliency map combined with super pixel feature classification is proposed.This method has high classification accuracy and can extract the natural contour line of the visual landmark.These two methods can effectively overcome the shortage of the traditional matching algorithms having high execution efficiency and meeting the lunar navigation and landing requirements.For the details of accurate navigation,this paper introduces the ORB-SLAM based navigation algorithm.However,this paper introduces an IMU fusion ORB-SLAM navigation algorithm.Traditional SLAM cannot estimate the real scale and is easy to be affected by noise during navigation.Pure IMU navigation algorithm has cumulative error problems which cannot release by IMU self.In order to solve this problem,visual SLAM system is used to act as auxiliary system of IMU,which means visual SLAM helps to release this error.In the aspect of nonlinear optimization,the data from IMU is set as the initial value for the optimization process.But when visual SLAM is lost,the cumulative error is only released through loop closure.At the end of this paper,compared with ORB-SLAM,the IMU fusion ORB-SLAM is much more stable and accurate.