Texture is the key character of remote sensing image. In this paper, the image texture was extracted by means of semivariogram. On such a basis, this study adopted the back propagation artificial neural network method to make classification by combining spectral feature with many sort of textures. The classification results were then compared with the results obtained by the maximum likelihood method. The results of the study have proved that the application of combined spectral features and textural measures based on the geostatistics and NN theory to the classification of the remote sensing image may improve the accuracy of image classification.
李小涛, 李纪人, 黄诗峰, 宋小宁. 变差函数和神经网络结合的遥感影像分类方法研究[J]. 国土资源遥感, 2006, 18(1): 18-21.
LI Xiao-Tao, LI Ji-Ren, HUANG Shi-Feng, SONG Xiao-Ning. A REMOTE SENSING IMAGE CLASSIFICATION
METHOD BASED ON GEOSTATISTICS. REMOTE SENSING FOR LAND & RESOURCES, 2006, 18(1): 18-21.