关键词: 撞击坑, 边缘清晰度, 评价算法, Sobel算子, 小波变换, 信息熵, 反向传播（BP）神经网络, 数字月球平台

Abstract: Impact crater is one of the most important geological structures on the surface of moon. By appraising the edge definition of impact craters in the CCD images of “CE-1”, some geological information can be inverted such as the rate of decay and the undulation of topography. Considering the gradient in spatial domain, the high-frequency component in frequency domain and the theory of information, an algorithm derived from image definition appraisal is presented, aiming at appraising the edge definition of impact craters. Altogether three algorithms are realized, respectively based on Sobel algorithm, wavelet transform and entropy of information. A kind of BP neural network which is adapted to the feature of impact crater is designed and results from different algorithms are combined as the inputs to the network, thus the final rank of the edge definition is achieved. The final results are loaded into the Digital Platform for Moon（DPM） to be displayed and make further analysis.

Key words: impact crater, edge definition, appraisal algorithm, Sobel algorithm, wavelet transform, entropy of information, Back Propagation（BP） neural network, Digital Platform for Moon（DPM）