Research On Rock Desertification Information Extraction And Geological Driving Factors Based On Multi-source Data

Rocky desertification has always been an important ecological problem affecting the sustainable development of Guizhou.The area of karst rocky desertification in Guizhou Province is wide,the terrain is steep and strange,the grade of rocky desertification is complete,deep and harmful.The rapid development of modern remote sensing technology,computer technology and artificial intelligence provides powerful support and possibility for accurate and rapid investigation and study of the present situation,distribution area,distribution disciplinarian and statistical analysis of the internal factors of rocky desertification in Karst areas.At present,the new deep belief network(DBN)model has not been used to extract the information of rocky desertification,and the relationship between geological factors and the occurrence and development of rocky desertification has not been quantitatively studied by taking geological driving factors as the entry point.Based on the project of "Construction of First-class Discipline of Ecology in Guizhou Province",this paper uses remote sensing data of three phases in 2000,2010 and 2017 as image data source,uses digital elevation model as topographic data source,and vectorizes geological map of southern Guizhou to obtain lithologic distribution data,establishes classification standard of rocky desertification grade in southern Guizhou,and adopts comprehensive analysis method,decision tree method and depth belief network(DBN)model.The information of rocky desertification in southern Guizhou is extracted by type method.On this basis,the dynamic evolution of rocky desertification in southern Guizhou is analyzed by using the difference method.The changing rules of rocky desertification in southern Guizhou between 2000 and 2010 and 2017 are obtained by statistics.The geological driving factors of rocky desertification in southern Guizhou are analyzed in depth with the help of GIS platform.Geological factors are fundamental to the occurrence of rocky desertification.Studying the relationship between rocks,karst landforms and geological structures and the occurrence and development of rocky desertification shows that the occurrence of rocky desertification is aggravated by long-term improper human activities in weak geological background areas.Therefore,it provides a reasonable reference for monitoring and controlling rocky desertification in southern Guizhou and karst mountainous areas of Guizhou Province.The main results are as follows.(1)Aiming at remotely sensed image denoising,a new denoising algorithm based on two-dimensional EMD and adaptive Gauss filtering is proposed on the basis of MATLAB programming.It is innovative and practical,and achieves better denoising effect.The algorithm is used to process remote sensing images of three phases in southern Guizhou.(2)According to the grading standard of rocky desertification in southern Guizhou,three methods are used to extract rocky desertification information in the study area,namely,multi-index comprehensive analysis method,knowledge-based decision tree method and depth belief network(DBN)model method.The depth belief network(DBN)model method is mainly studied to extract rocky desertification information from remote sensing images,that is,to explore the information extraction method of artificial intelligence.According to field survey data and Google map,the overall classification accuracy of rocky desertification information extracted by comprehensive analysis method is 71.05%,Kappa coefficient is 63.34%;the overall classification accuracy of rocky desertification information extracted by decision tree method is 54%,Kappa coefficient is 41.74%;the overall classification accuracy of rocky desertification information extracted by depth belief network(DBN)model is 77.17%,Kappa coefficient is 71.20%.The deep belief network(DBN)model method has the best effect and achieves the desired purpose,which has a positive effect on promoting the application of artificial intelligence in remote sensing.(3)In the process of extracting the information of rocky desertification,the remote sensing image characteristics of limestone,dolomite and limestone-dolomite interbedding are deeply analyzed by Landsat 8 remote sensing image,and the remote sensing image characteristics of three carbonate rocks in different tectonic regions are also analyzed.The study of these image features not only provides a reference for the study of karst landforms,but also provides relevant reference materials and data for the development of basic geology and applied geology.(4)Using deep belief network(DBN)model to extract the information of rocky desertification in 2000,2010 and 2017,and calculating the difference to obtain the distribution maps of rocky desertification change degree between 2000 and 2010 and 2017.By means of GIS statistics,it was found that rocky desertification in the study area was aggravated and deteriorated in 2000 and 2010,but the rate of aggravation was decreasing year by year and rocky desertification in some parts of the study area was found.The situation of rocky desertification has been improved greatly from 2010 to 2017,and the area of improvement of rocky desertification is much larger than that of aggravation.(5)According to the results of rocky desertification extraction from remote sensing images of three time phases,the correlation between graded distribution of rocky desertification and geological driving factors(rock formations,karst landforms and geological structures)in the study area is analyzed.The results show that the occurrence rate of rocky desertification in limestone area is the highest and the situation of rocky desertification is serious,and the lowest in the area where carbonate rocks and clastic rocks interbedded.The occurrence rate of rocky desertification in the interlayer jointed structure-karst area is higher than that in the conjugate jointed structure-karst area,but the distribution area of rocky desertification in the interlayer jointed structure-karst area in southern Guizhou is 5836.25 square kilometers,much smaller than that of rocky desertification in the conjugate jointed structure-karst area.The distribution area of the structure-karst area is 20374.83 square kilometers.(6)the geological background factors of rocky desertification in the karst area of southern Guizhou will not change and will remain stable for a long time.Distribution characteristics of various geological background factors and their relationship with the occurrence and development of rocky desertification are very clear in spatial distribution and expression,which is helpful to carry out the key monitoring and guiding control of rocky desertification in the study area.