Visual Navigation During Asteroid Landing

Asteroid exploration,as an important branch of deep space exploration,has great scientific significance.Through the exploration of asteroids,we can understand the origin of the solar system,find resources that can be used by human beings,and find ways to prevent asteroids from hitting the earth.In asteroid exploration missions,improving the accuracy of the navigation system can avoid obstacles encountered during landing and guide the precise landing of the probe.During the descent and landing phase of the probe,there are great differences between the image of the planet surface collected by the traditional visual navigation method and the two-dimensional navigation map carried by the detector itself in scale,angle of view and illumination direction,which will cause difficulties in feature extraction and matching,and then affect the navigation accuracy.Based on the background of the descent and landing of the probe on the asteroid Eros433,a new visual navigation method is proposed in this paper.The three-dimensional model of the target planet is established according to the information collected during the flight phase of the probe.The navigation map is generated in real time according to the position and attitude of the predicted detector.The navigation map generated by this method and the target planet table captured by the navigation camera.The difference between face images is small and the feature changes little,which reduces the difficulty of feature extraction and matching.Aiming at the navigation method proposed in this paper,the navigation map is generated according to the three-dimensional model of the target planet and the position and orientation information of the predicted detector.The real image captured by the navigation camera is simulated by the image generated by slightly errors of the position and orientation of the detector.After the feature points extraction and matching of the navigation map and the real image are carried out,and the mismatching is removed.The correct matching rate of feature points is 98.74 %.The pixel coordinates of the correctly matched feature points in the navigation map are transferred to the navigation filter as the observation.Because the system state equation of the descent and landing phase of the detector is non-linear,the extended Kalman filter is used to filter the navigation.Finally,the more accurate navigation parameters of the detector are estimated.