| The Loess Plateau is a unique natural geographical unit,its soil complex,special climate, terrain fragmentation, fragile environment, ravines. So this region developed a different shape, different sizes, different proportions of the special landscape loess sinking hole. Sinking hole has distinct cave features, widely distributed and harmful,and is mostly distributed in areas with strong collapsibility of loess strata. Some of them are scattered on the surface,some are beaded or honeycomb. Its development is closely related to soil erosion and other geological disasters, but also endanger the safety of all kinds of projects and people’s lives and property. The method of field investigation is high reliability, but it is time-consuming and laborious. With the development of surveying and mapping,remote sensing,remote sensing data and DEM data accuracy gradually increased.These provides a new idea and method for the study of micro geomorphology.Therefore,how to quickly identify the loess sinking hole disaster has become an urgent problem to be solved.This paper takes the typical area as the study area in Dingxi of Gansu Province,based on Google and DEM image data as the main data source, after complete quantitative multi-scale segmentation, using object oriented method to automatically extract sample area of loess sinking hole. Finally, combined with field investigation,this paper analyzes the distribution characteristics and development characteristics of loess sinking hole in the study area:1. Compared with the traditional qualitative multi-scale segmentation method,this paper focuses on three parameters of multi-scale segmentation from the quantitative point of view. Firstly, By comparing all the applicability of the method,combined with the characteristics of this data, choosing the scale where brightness mean standard deviation is maximum as the optimal scale. Secondly, Refer to the Maximum area method, using the constant interval of maximum area corresponding ground patch as the optimal interval of shape and compactness. Realizing quantitative selection of scale, shape and compactness. According to the segmentation parameters obtained, realizing multi-level and multi-scale segmentation of image.2. Using object oriented method to realize automatic extraction of loess sinking hole, this paper extraction sinkhole 267. The distribution of sinkhole is displayed objectively. Comparison of the accuracy and field survey, found that the extraction results coincide 241, its accuracy reached 92.3%. In addition, this method is used to extract typical area in Dingxi,930 were extracted. In this paper,the automatic extraction of the region of loess sinkhole provide technical support for the extraction of sinkhole and other typical micro topography in Longxi Loess Plateau region.3. This paper is based on the DEM data analysis and field investigation to explore distribution of loess sinking hole. It is mainly distributed in the slope of 8° to 25° in the edge of the terrace edge, the concave slope below the edge of the Loess Plateau hilly and gully, so it is most distributed in the gully region. It is partly located on the concave slope of 25° to 60°; The loess sinking hole is often distributed in groups, individual distribution less, some showed beaded, some gourd shaped, and exhibit certain combination relations. |
PDF Full Text Request