A fast auto recognition algorithm for lunar terrain in wavelet domain

With the development of the science and space technology, human collected a lot of image data about the lunar terrain. Since the middle of the 20th century, human began landing experiment to the Lunar. So the study of lunar terrain had become a hot topic in recent years. This paper proposed a new automatic recognition algorithm. We introduced the Wavelet domain into the lunar terrain recognition. This new algorithm did Wavelet Transform to Lunar CCD data first, then according to the difference of DWT components, we chose different features to form feature vector. Then we normalized the feature vector, finally we used K-means to cluster in vector space. In order to compared with the existing algorithm fairly, we chose four typical areas: ‘H010’, ‘SI’, ‘Crisium’ and ‘W4’ as testing areas, we compared the recognition rates and Cohen's kappa coefficients with three previous algorithms. The results show that the new algorithm has satisfied results with more faster processing speeds.