Research On Registration Of Large-scale Transformed Images Of Lunar Rover Traveling At Large Distances

In Chang'E-4 lunar exploration mission,Yutu 2 rover is planned and controlled by teleoperation.The rover usually travels between 6 and 9 meters in order to make the distance of each planned travel more extensive,which makes the images taken by the Yutu 2 rover differ saliently in scale,rotation and translation and the overlap areas of these images have small size.Notably,the resolution and morphology of the overlap area are very different,which brings great difficulties to image matching,thus hindering automatic visual positioning between adjacent stations.To overcome these difficulties,this thesis analyzes the influence of spatial resolution ratio on image matching from the perspective of scale-space difference,and build the mapping relationship from the length of photographic ray and the angle between the ray and the surface normal to the spatial resolutions of each image pixel,which quantificationally describes the overlaps and scale transforms of images in the condition of station distances,imaging distances and lunar surface inclination angles.With the images acquired by Yutu 2 rover in Chang'E-4 mission,the proposed method is validated by experiments.The specific research contents are as follows:To begin with,from the perspective projection transformation,we consider the normal vector of the observation point on the lunar surface and other factors,deduce the influence of the unit pixel change on the actual spatial position,and establish a general pixel spatial resolution model.After analyzing the imaging relationship between adjacent stations,we derive a general model of the ratio of pixel spatial resolution.Besides,we detailedly discuss the variation of image scale and the rate of feature matching in terms of station distances,imaging distances and lunar surface inclination angles or their combinations,and develop the mechanism to choose the equivalent minimal factor subset,so that the influence of the combination of different factors on image scale and feature matching can be analyzed comprehensively.Thus,the influences to localization of rover from station distances,imaging distances and lunar surface inclination angles are evaluated and established,which could be utilized to guide a more compelling image registration and visual localization.Furthermore,the proposed model is evaluated by using the station data of Yutu-2 rover collected in the first and second lunar days.With a detailed analysis of the influence of thepixel space resolution ratio on the rate of image matching,the influence cases are classified into four types.A general rule is given for describing image matching rate under different stations,and the constraints to determine the distance and location selection of adjacent stations in an actual task are presented,which paves the way for the improvement of the visual positioning level of the subsequent extraterrestrial objects exploration.