Study On Optical Navigation Of Asteroid Lander Under Poor Illumination

Asteroid exploration is of great significance to human beings.Deploying a probe to land on the asteroid is the basis for deep exploration of the asteroid.For an asteroid lander that cannot rely on radio positioning from earth,autonomous navigation is the key technology to determine whether the mission can be completed.Among the many navigation methods,optical navigation method has the advantages of light and reliable equipment,low energy consumption,and strong applicability in the weak gravitational environment of asteroids and the light without atmospheric shielding.Due to the fast rotation speed of the asteroid,the light on the asteroid’s surface will constantly change.In order to realize the autonomous navigation of the asteroid lander using the optical navigation method in the landing,it is necessary to solve the problem that the optical navigation is sensitive to light.Based on the background of the optical autonomous navigation of the asteroid lander,this paper studies the light sensitivity issue of the optical navigation method,proposes effective solutions from the perspective of image preprocessing and feature extraction,and improves the robustness of the optical navigation method to light.The main contents of the paper are as follows:First of all,the optical navigation methods were studied,according to the characteristics of the asteroid landing mission,an optical navigation method for vector measurement based on matching with feature library are studied.In order to reduce the measurement error of the optical navigation and enhance the robustness of navigation system for illumination,we studied the inertial navigation method,and an optical inertial integrated navigation system is designed.Secondly,the feature extraction and feature matching methods,which are important steps in optical navigation,are studied,and the MET feature extraction method,k-means feature extraction method and Hu moment feature region matching method are introduced.Because in optical navigation,feature extraction and matching are susceptible to adverse light enviroment,aiming at this problem,the light environment near asteroids were thorough analyzed,and we established the reflection light model based on the of the sun Angle.After studied the influence of illumination on the image,an image model of adverse light based on the Angle of solar irradiation is established,which can provide experimental conditions to prove the illumination robustness of optical navigation methods.Then,based on the analysis of the asteroid lighting environment,gamma transform method based on the sun Angle and improved image detail enhancement method based on gaussian pyramid is proposed,these two methods are used as an image preprocessing step in optical navigation.We used feature extraction method proved the image preprocessing method can effectively improve the feature extraction methods of illumination robustness.Finally,in order to solve the problem of the traditional feature extraction methods that have low accuracy and sensitive to light,we put forward an asteroid’s surface feature extraction method based on the deep learning which uses YOLO V3 neural network as the foundation structure and annotated images of asteroids and other planets as the data set.The poorly illuminated image generation method was used to enhance the training set,and a feature extractor with strong illumination robustness were obtained after the neural networks were trained using both enhanced and unenhanced data sets.The comparison experiment proves that this method is more accurate than the traditional feature extraction method,and it can also extract features stably in the image with poor illumination.The feature extraction method is verified by the navigation sequence image with poor illumination,and it is proved that the feature area extraction method based on deep learning can guarantee the completion of navigation task under the condition of poor illumination.