Study On Boundary Extraction Of Kumtag Desert Based On Remote Sensing Technology

China is vast in territory and abundant in natural resources.As a natural resource,the desert has also received great attention.With the continuous development of remote sensing technology,its technology is increasingly mature and gradually applied in various fields.Remote sensing technology is also used in the study of the extraction of the desert boundary line.Taking the Kumtag desert in Piqan County as the research object,remote sensing image as the data source,the boundary line of Kumtag desert in Piqan County has been extracted.This work can provide the basis for the completion of the cadastral survey and definitive registration of natural resources in the Kumtag desert region of Piqan County.At the same time,it can provide the basis for formulating policies on ecological environment protection and rational utilization of natural resources.In this paper,the remote sensing images of August 1,2020 were classified by using the maximum likelihood,Support vector machine and super pixel algorithms.The classification results for maximum likelihood and Support vector machine are based on multi spectral remote sensing images and texture features respectively.For the super-pixel algorithm,an improved super-pixel algorithm based on compactness factor is adopted.It can adjust the weight of compactness factor reasonably.Finally,Edge Point Connection and NDDI are used to extract the desert boundary.At the same time,this paper also completed the mapping work of Kumtag desert boundary by the method of manual mapping.Based on the results of artificial surveying,the area of the desert obtained by artificial surveying is compared with the area of the desert obtained by the three methods.The boundary accuracy of the maximum likelihood,the Support vector machine and the improved super pixel algorithm are obtained.The results show that the overall accuracy of the improved super pixel algorithm is 93.50%,the Kappa coefficient is 0.88,the support vector machine is 90.80%,the Kappa coefficient is 0.83,the maximum likelihood algorithm is 84.30%,the Kappa coefficient is 0.75.This shows that the accuracy of different classification methods has a certain difference.The total accuracy of the three methods is above 80%,the kappa coefficient of improved super-pixel algorithm and support vector machine are both higher than 0.80,and the maximum likelihood is not up to 0.80.The improved super-pixel algorithm and support vector machine both meet the classification accuracy requirements.The classification accuracy of maximum likelihood in remote sensing information extraction is inferior to that of the other two methods,and there is room for improvement.The area obtained by artificial mapping method is 2167.00 km~2,the area gained by maximum likelihood method is 2272.30 km~2,the error value is 105.30 km~2,the relative error is 4.86%,the area gained by support vector machine is 2232.70 km~2,the error value is 65.7 km~2,the relative error is 3.03%;The area of the improved super-pixel algorithm is 2226.60 km~2,the error value is 59.60 km~2,and the relative error is2.75%.But the area obtained by the improved super-pixel algorithm is the closest to the area obtained by manual surveying method,which proves that this method is more reliable in the desert boundary extraction.From the above analysis,it can be concluded that the multi-spectral remote sensing image based on the medium resolution can be used in the desert boundary extraction in the area of county,whether from the point of view of classification processing or boundary extraction.The precision of the three methods is above 80%,and the relative error is less than 5%.Therefore,the remote sensing technology based on the Kumtag desert boundary extraction can meet the accuracy requirements,can provide a certain extent support for the work of definitive registration of natural resources.