Texture of wetland in single-band and single-polarization ENVISAT ASAR image

Laixiang Jiang,Jie Zhang,Yi Ma

2010-01-01

Abstract:Texture is an inherent property of virtually all surfaces. It contains important information about the structural arrangement of surfaces and the relationships with the surrounding environment. Image texture is a kind of important feature for object recognition. The paper has taken the ENVISAT ASAR image in the Yellow River Delta as an example, and analyzed four typical wetland types (aquaculture, salt field, tidal flat, and reservoir) and two other land-use categories(farmland and sea) based on gray level co-occurrence matrix(GLCM). We ascertained the factors of GLCM on SAR wetland image such as orientation, displacement, quantization, and window size. The optimum orientation was average of the four orientations of 0°, 45°, 90° and 135°. The optimum quantization in these texture parameters was 64-level. The optimum window size was 13×13. With regard of the displacement, the optimum was 6. The textural analysis results indicated that reservoirs had lower gray value, marked boundary and the difference of texture curve between it and other wetland types; salt field and aquaculture had similarity on spatial structure and gray value, so their texture curve were close. Because the image had not been filtered before texture analysis, there were noise, the texture of sea was close to that of tidal flat. Farmland has a feature of higher gray value, complex spatial structure and confused orientation, so it was easy to distinguish it from others, although its texture feature was close to others.