中文摘要：

黄土沟谷是黄土地貌发育演化研究的重要切入点,可分为继承性与水蚀性两种类型。其中,水蚀性沟谷是构建区域性地表过程模型的重要参数。然而,利用现有的基于数字高程模型（DEM）水文分析方法,无法准确识别水蚀性沟谷。基于黄土坡面流线转折特征,系统性提出一套利用沟沿线约束的黄土水蚀性沟谷识别方法。在宁夏回族自治区固原市刘家沟流域实验结果表明：该方法可准确区分两种类型沟谷;提取的水蚀性沟谷在偏移距离与沟谷密度指标方面,与目视解译结果对比,精度分别为84.01%,96.98%。

英文摘要：

Water erosion is one of the key momentums of shaping the external appearance of modern loess landform. Studies on waterworn gullies are not only significant in loess digital terrain analysis but also useful for recognizing evolution laws of loess landform. As a result,accurately identifying and extraction of waterworn gullies have always been the vital basic and critical part in quantitative research on the Loess Plateau.However,a majority of loess gullies are usually combinations of inherited gullies and waterworn gullies on southern Ningxia,eastern Gansu and western Shaanxi Loess Plateau,which makes great difficulty in distinguishing the waterworn gullies by using of hydrological analysis modules. Severe deviation would be caused in research of calculating core topographic indices（ gully density） if we cannot make a distinction between the two types of gullies. To improve the accuracy of waterworn gullies extraction by using of hydrological analysis modules,this paper proposed an extraction method under constraint of loess shoulder-line.We select the Liujiagou watershed as test site because it belongs to hilly Loess Plateau region of China,and there are vast combinations of inherited gullies and waterworn gullies. Our original test data are 5 m-resolution DEM and match Quick Bird images with 0. 61 m-resolution. The DEM was derived in 2006 and it meets the national standards of China; ortho-rectification of Quick Bird image was conducted by using of 5 m-resolution DEM.Our algorithm includes four steps： 1） extraction of original gullies： In this part,raster gullies are extracted by using of hydrological analysis module by support of GIS software and transformed into vector format. After that,vector gullies are compiled based on high resolution image. 2） Generation of qualified shoulder-lines： Based on the fact that shoulder-line is locating at the slope breaking points of stream line on the loess slope,a new generation method is designed for detecting of shoulder lines. After transforming the extrac