Research On Automatic Extraction Of Erosion Gully In Black Soil Region Of Midwest Jilin Province Based On Multi-Source Data

The black soil area in the central and west part of Jilin Province is the most fertile land resource in Jilin Province.However,due to human activities and some changes in the environment,soil erosion is serious in black soil area.Among them,the effect of gully erosion is the most important.And the erosion gully caused by gully erosion seriously damages the integrity and area of cultivated land.Therefore,it is very necessary to extract and monitor the erosion gully.In the research of the erosion gully,most of them are based on the method of visual interpretation,and the adopting of automatic extraction for exploring is less.Based on two kinds of resolution remote sensing images of GF-1 Satellite and combined with the geographic data,the black soil area in the central part of Jilin Province was taken as the research sampling region in this paper.The research exploring on the automatic extraction of the erosion gully was carried out from the angle of big and small region respectively.A set of automatic extraction scheme was designed in this paper,which provided experience for the research in the relevant filed and technical support for soil and water conservation in black soil region of Jilin Province.The main research content and conclusion in this paper were as follows:(1)The concept and classification system of the erosion gully were described,and the development mechanism and distribution characteristics of the erosion gully in the black soil area of northeast China were analyzed.The current research status on the erosion gully at home and abroad was summarized,and the research thought on the automatic extraction of the erosion gully in the black soil area in Jilin Province was proposed.(2)Based on SRTM1 DEM,some data such as slope data,catchment cumulant data,river network data and graded gully line data were extracted to conduct the automatic extraction research on the erosion gully with the remote sensing images.(3)Based on the object-oriented global optimal partitioning algorithm,the optimal partitioning scale of the research sample region was achieved.The partitioning result showed that the scale achieved by the global optimal partitioning algorithm was not the optimal scale of the erosion gully.The optimal partitioning according to the erosion gully can be achieved by the visual comparison based on the scale achieved by the global optimal partitioning algorithm and through comparing its neighboring scales.(4)The feature space optimization method for constructing forest model was explored,that is,the extracted feature space was optimized based on the importance of the feature.The result showed that the spectral feature made most contribution to the model,followed by texture feature and finally geometric feature.(5)Based on two-meter resolution data of GF-1 Satellite,the object-oriented analysis method and the random forest model were used to realize the automatic extraction of the erosion gully in the research sample region of Jialong Town and Xiying Town in Jiutai District of Changchun as the scale of the small region,and the extraction result was proofed by using the gully line.The extraction result was analyzed carefully.Among them,the precision of Longjia Town was higher than that of Xiying Town,and the overall precision of their erosion gully was 77.95% and 73.60% respectively,which showed that the geographic feature had an impact on extracting the precision.(6)Based on sixteen-meter resolution image of GF-1 Satellite and the geographic data derived by DEM,an automatic extraction method for the erosion gully in the black soil sample region facing the big region was designed.This method was also designed based the object-oriented analysis method and the random forest classification model,but integrated with remote sensing images,slope data and catchment cumulant data in the data.Based on this method,the automatic extraction of the erosion gully in the big region of the black soil area in Jilin Province was achieved,and the overall extraction accuracy of the erosion gully was 73.78%.(7)The total number of erosion gullies extracted from Longjia Town sample area is 4124,with relatively uniform distribution.Large erosion gullies mainly distribute in the southeast and northwest;Xiying Town sample area mainly distributes in Hilly areas.There are more erosion gullies extracted than Longjia Town sample area,with a total of 4812 gullies,especially in the middle and eastern mountainous areas.A total of 59 024 stabilization gullies were extracted from the large-scale sample area,which were distributed densely in the hilly areas of the central and Western regions.