This paper deals with the techniques for improving image classification by extracting geometric attributes of objects. In this paper, we first introduce one kind of symmetry transform called directional symmetry transform, which can provide more geometric information about objects. Based on it, we present one algorithm for improving image classification by determining scale factors relating to supervised learning. The results show that the number of the mixed pixels can be efficiently reduced by using the proposed method.
关泽群, 刘继琳, 崔卫红 . 利用地物的几何属性改进影像分类效果[J]. 国土资源遥感, 2001, 13(1): 31-35,65.
GUAN Ze-qun, LIU Ji-lin, CUI Wei-hong . EXTRACTION OF GEOMETRIC ATTRIBUTES OF OBJECTS FOR IMPROVING IMAGE CLASSIFICATION. REMOTE SENSING FOR LAND & RESOURCES, 2001, 13(1): 31-35,65.
[1] 周杰.基于方向对称变换的人脸定位方法【J】.电子学报, 1999, 27(8):12-15.[2] 孙家柄,刘继林,李军.多源遥感影像融合【J】.遥感学报, 1998, 2(1):47-50.[3] Preparate F P,Shamos M I. Computational Geometry: An Introduction. Springer-Verlag,1985.