A Study On Methodology Of Soil Conservation Practices Extraction Based On Texture Features Of Hi-resolution Remotely Sensed Imagery

The hill and gully region of Loess Plateau suffers seriously from soil erosion and have been taken as a national key-harnessing region for the water and soil conservation. It is a fundamental task to monitor soil and water conservation dynamically in the region. In terms of extraction of soil and water conservation practices, moderate-resolution remote sensing images, sepecially Landsat TM image, were taken as main data source in the previous studies. Limited by the spatial resolution, it is difficulty to extract small pieces of the distribution of soil and water conservation practices, such as terrace, orchard, grass, plantation land and dam land based on these images. With the increasing availability of high-resolution remote sensing images, soil and water conservation monitoring by high-resolution remote sensing images now is not only important, but also very urgent in the hilly and gully region of Loess Plateau.Taken Ju Yuan'Gou drainage in Suide County and Ma Zhuang watershed in Baota District of Yanan city of northern Shaanxi province as test areas, and 1m IKONOS image and 2.5m SPOT-5 image as data sources, based on the character of energy superposition from space to frequency by Fourier transformation, terracing and orchard were extrcted by using classification method of the ratio of the maximum orientation in frequency (combination spectral information). Then, the differences were discussed between the new method and two traditional supervised and unsupervised classification in terms of extraction of soil and water conservation practices. At last, results of supervised classification and unsupervised classification were contrasted. The main results of this study are:1) Used classification method of the ratio of the maximum orientation in frequency combined spectral information, the information of terrace and orchard from IKONOS image and SPOT-5 image was extracted. The total accuracy can be achieved more than 78.21%2) In the aspect of the extraction of terrace with big pieces, 10m in breadth and big difference in space, the results of SPOT-5 is satisfactory. The total accuracy can be achieved more than 78.21%. IKONOS image is good at the extraction of terrace with 3-7m in breadth. The total accuracy can be achieved more than 81.36%.3) The method based on texture features has higher accuracy than supervised classification and unsupervised classification. The terrace extraction result of supervised classification from SPOT-5 image is similar to unsupervised classification, but supervised classification result was better than unsupervised classification result in the orchard extraction. While for terrace extraction, supervised classification has better result than unsupervised classification. The difference of total accuracy between two methods based on IKONOS image is more than 12%.The extraction of soil and water conservation measures with texture features was studied in this paper. The objective of this research is to provide reference to investigation on regional soil erosion and benefit appraisement.