Abstract: With the analysis of grey and texture features, a key-degree index of classification features was presented to solve target classification in synthetic aperture radar (bAR) images. The key feature, the hypo-key feature, the non-key feature and the feature key-degee were defined by the concept of the key coefficient. The high key-degree features were selected from the feature statistic of grey and tex?ture models, for example, mean and variance from grey model, and angular second moment, contrast, homogeneity and correlation from texture model. Aiming at multi-class and multi-feature of SAR im?age classification, we classified the samples by the principle of minimum distance on the basis of con?structing feature vector and defining vector distance. To improve furture computational speed and ae- scribe correctly the feature, we introduced a window approach to the classification of SAR images. The simulation and the calculation results show that the approach is effective.

Key words: information processing , synthetic aperture radar image , feature extraction , target classifi?cation , image processing