| Sandy cropland is a distinctive dryland farming form in northwest China, which is the wisdom of the people in the Northwest. It can adapt to the natural conditions of arid climate^poor soih scarce precipitation and intense evaporation, which limit agricultural production. As the largest sandy cropland base in China, Zhongwei sandy cropland around Xiangshan has planted widespread Xisha melon. The products are exported to Hong Kong, Macao and the economically developed eastern regions, and it forms green product industry here. Thus, studying the spatial distribution and variation of zhongwei sandy cropland has very important significance. As a reliable resource, which has the advantage of updating fast, acquiring data easily, covering widely and application technology maturely, remote sensing provides an effective way for sandy cropland study.With the series of Landsat images and the object-based image analysis method, the paper extracts sandy cropland from five periods of images from1992to2013by establishing rule sets, and analyses sandy cropland’s spatial distribution pattern and time variation in more than twenty years. The main conclusions of the paper are as follows:(1)The basic processing unit of object-based image analysis method is objects, which is internally homogeneous and heterogeneous between adjacent objects. And It can reflect the class surface unit better. The wide use of texture, shape, size, context and ancillary GIS data in analysis can achieve efficient extraction of sandy cropland. It is able to perform better in the integrated analysis with GIS data by exporting segmentation and classification results in vector form.(2) The paper uses discrepancy measures to evaluate segmentation parameters sets, which considers both geometric discrepancy and algebraic inconsistency. By this method, the optimal segmentation parameters of sandy cropland are as follows:scale factor is60, shape factor is0.7, and compactness factor is0.8. The experiment indicates that the results of this segmentation parameter fit real sandy cropland best.(3) Sandy cropland and desert shrub, settlement place are very easy to mix in the image. This paper establishes sandy cropland’s rule set according to the principle of removing irrelevant field step by step from easy to difficult. Desert and water are removed by the wetness feature derived from Tasseled Cap. Farmland and forest land are extracted by the near infrared ratio feature. Integrating temperature, slope, shatian index (NDSI) and newNIRratio feature, shatian is distinguished from desert shrub by the membership function of the features. The settlement place is got with the adjacent object boundary ratio feature. The result shows that the extraction accuracy of this rule set can meet the requirement.(4) In the past twenty years, Sandy cropland has developed rapidly in Zhongwei city. In1990s, sandy cropland takes on linear distribution in Xiangshan town. At the beginning of twenty-first century, shatian cropland expands to Xingren town and Hanjiaoshui town around Xiangshan Mountain. About2007, shatian begin to distribute in Nanshantaizi. Now, shatian has spread all over the Xiangshan Region with wide spatially distribution, and take on radially planar distribution.(5) During the other period, Sandy cropland changes slowly from1999to2002, while changing quickly or dramatically. The area of sandy cropland has increased about45thousand hectares during the research period. Over ninety percent of new sandy croplands are formed on the basis of the existing desert scrub, and a small portion occupies the infertile farmland. The former turns waste into treasure of land, and can reduce desertification. The later increases the land value.(6) In the recent twenty years, the rapid development of sandy cropland is benefited from the superior natural resource in zhongwei city. And the main driving factors are the guide of nation and government policy, the driver of economy and traffic, and the population migration and growth. |
PDF Full Text Request