An improved recognition algorithm for lunar terrain based on CCD image

With the development of space technology, topographic study of celestial bodies becomes increasingly important. In order to better carry out geomorphologic analysis and landing site selection of celestial bodies, the terrain classification becomes particularly critical. This paper provides an improved algorithm and proved its better in identifying the lunar mare area and the highland area of the CCD images with four features used in k-means clustering. We chose two typical areas: 'H010' and 'SI' areas of lunar terrain to research. And the result of the improved algorithm is analyzed from two different block size with different number of testing points. And also the whole recognition rate and Cohen's kappa coefficient are both better than the result of previous algorithm in using DEM or CCD data. Especially in the 'H010' area, the average whole recognition rate is 91.4325%, and the average Cohen's kappa coefficient is 0.813. In this paper, we use the CCD data of Chang'E-1 downloaded from The Science and Application Center for Moon and Deep space Exploration. And we improved the algorithm which based on Jiang's method (10) , we choose the suitable block size and set four features to distinguish and recognize the lunar mare and highland, and we compared two results from different block size. The result has been compared with the geological data published by USGS which has shown that the improved algorithm does well in blocking image for terrain classification.