Research On Classification And Extraction Of Impact Craters On Moon Surface Based On Morphological Features

With the advancement of deep space exploration technology,the moon's important position as the closest satellite to the earth has become more prominent.The impact craters that are widespread on the surface of the moon are windows into studying the moon.Through these impact craters,important information such as the formation and evolution history of the moon can be learned.Impact craters are characterized by a large numbers and diverse forms,therefore,reasonable classification schemes and effective detection methods are of great significance to the development of various research works on the moon.Crater formation periods and evolutions processes vary,hence their expressed texture characteristics are usually different.So,this study first classified the impact craters according to their morphological characteristics,and proposes the division of the moon surface impact craters into seven types: simple craters,Floor-fractured craters,central peak craters,rough bottom craters,incomplete impact craters,residual craters in lunar sea and circular basins.Then clarified the characteristics and formation reasons of each craters type,and then compared and explained the similarities and differences between categories.Finally,the size distribution range and geographical distribution position of each impact crater type were statistically analyzed,and a relatively complete classification result of impact crater was estimated.Existing impact crater detection methods are mainly aimed at small-size impact craters with simple shapes.These algorithms cannot achieve the expected detection effect for impact craters with large diameters and complex textures.Based on the circular features shared by the impact craters,this paper fully considered the diversity of the expression of the impact craters,and proposed an automatic extraction method of impact craters based on the difference distribution.This method uses image features and machine learning algorithms as its main tools.First,HOG features were used to describe the characteristics of the impact crater,and then the TrAdaboost classifier was used for automatic detection.In view of the different degrees of difference between the impact craters,this employed relative entropy to calculate the difference between each data,select the data within a certain tolerance range as the training sample,and increase the number of samples while ensuring the quality of the sample,in order to improve the accuracy of the classifier and the detection effect of the algorithm.Experiments conducted show that the proposed crater detection method has a good performance and high detection accuracy,and can meet the requirements for large and complex crater(s)extraction.Finally,the neural network principle was applied to automatically classify the complex impact craters according to the above classification method,and good classification results were obtained.