Automatic Extraction Of Valley Area And Geomorphological Analysis On The Loess Plateau Based On High Spatial Resolution Images

The valleys and canyons crisscross the Loess Plateau,and loess is especially susceptible to erosion by effect of water.Gully erosion is the main type of soil erosion in the Loess Plateau.The Loess Plateau is divided into two types of geomorphic units:inter-gully area and gully area,which are bounded by the gully shoulder line.The gully shoulder line is also the most active part of the eroding gully.The extraction of a gully shoulder line and division between the inter-gully area and the gully area are the basis technique in monitoring the development of the gullies and mastering the mechanism of soil erosion information in gully areas of the Loess Plateau.The automatic extraction technique presented in this paper aims to combine high-resolution remote sensing data and GIS methods.The varieties of automatic extraction methods for valleys,based on commercial images,Google images and terrain data,were designed by taking Xifeng District,Wuqi County and Jingde County as research areas.By comparing the extraction precision,the characteristic system of the gully geomorphology is further established,and the difference of erosion area in the Loess Plateau is analyzed.The key results are as follows:(1)When extracting gully area in the Loess Plateau Gully Region,the extraction accuracy of the nearest neighbor classification method,with spectral data as sole input data,is higher than the threshold classification method;Using threshold classification method.combining with spectral data and Terrain data,can improve the accuracy,the correct rate is 99.4%,and the extraction accuracy of this method with only spectral data is the lowest to extract the gully area.(2)When extracting the gully area in the Loess Hilly Gully Region,the extraction result of the nearest neighbor classification method,with commercial spectral data as sole input data,has low accuracy.The commercial image also can be replaced by the histogram stretching processed Google image.Combining with the region growing algorithm,designing the object-based classification rulerset,mapping gully area from Google image.This semi-automatic extraction can be achieved with high precision,the overall accuracy is 87.65%,and the Kappa coefficient is 0.75.(3)Based on terrain data and the automatic extraction result of the gully area and inter-gully area,geomorphological analysis was conducted on the Loess Hilly Gully Region.In the first sub-region,the shoulder line density is 18.08 km·km-2,the gully area average slope is 38.29°,and the approximation degree of basin boundary is 2.07.There are more erosion gullies,steeper gully slope,which means the development degree of gully area is higher.In the second sub-region,the shoulder line density is 7.36 km·km-2,the gully area average slope is 35.84°,and the basin boundary approximation degree is 1.55.There are more complex gully shoulder line and more gradual gully slope.