Spatial Features Analysis And Classification Of Landforms In Shandong Based On DEM Image Texture

Shandong area has a superior natural geographical environment.Through the comprehensive action of the earth and external forces,it has formed different types of landforms,such as plain,hilly and mountainous area,and then has constructed diverse regional landform types.Systematic acquisition of regional landform feature objects,concrete analysis of landform feature parameters,and systematic fine classification of landforms are the research hotspots in geoscience field.Landform morphology is the apparent reflection of the differences of different landform features,while the accurate classification of landform is the key step to describe the landform features of a specific region and to understand the formation.process of its internal landform.At present,most of the researches on landform features are based on the calculation of topographic factors to mine the quantitative law of local landform features,but the ability to grasp landform spatial features is insufficient,and the accuracy of identifying or classifying landform type is low.In view of the fact that nowadays the effective macro landscape analysis methods are lacking,this article start from the perspective of texture analysis,and then introduce the concept of image texture to the research field of digital terrain analysis technology.Using the theory of gray level co-occurrence matrix texture analysis method to quantitative extract the terrain texture feature value,qualitative analyze the formation mechanism of the landforms,and illustrate the changing rule of the macro landform space features,all of those are likely to achieve new breakthrough in the field of landform type classification.In this study,terrain texture in Shandong area is taked as the research object.Based on the multi-scale DEM data,gray level co-occurrence matrix method is adopted,and the relevant theories of regional geomorphology,digital terrain analysis,image processing of remote sensing and mathematical statistics are integrated to carry out a systematic and in-depth study on the texture feature parameters,macro landform spatial features and landform classification.The main work of this paper is described as follows,and has obtained the corresponding research results:(1)The texture feature values of texture measure are extracted from DEM data by using the method of gray co-occurrence matrix,and the influence of different calculation parameters on the texture feature value is analyzed.By virtue of the difference of texture feature value and the distinguishability in different landform types,the optimal parameters of direction,gray level,and texture measure are determined;while the optimal parameters of texture analysis window are determined by using the stability of variation coefficient.At the same time,DEM data with different resolutions are utilized to verify the consistency of feature expression scale,which proves that the optimal value of texture feature parameters is scientific.(2)On the basis of the optimization of texture feature parameters,specific analysis on texture feature values is performed,then the DEM comprehensive texture factor has been put forward,further quantitative analysis of landform space features and change law relied on texture are studied,and qualitative description of landform distribution pattern in Shandong area is achieved.Compared with.the existing landform map,it was found that DEM comprehensive texture factor that reflect the features of landform space distribution has high degree with landform type by one-to-one,which verify that the textures and landform types exist a strong coupling relationship,mapping a mechanism between texture and landform types.(3)According to the mapping mechanism between texture and landform types,texture is selected for landform classification in different scales.The distinguishing measure of different landform types is the difference of texture feature values.Through experiments,mean,variance,and contrast texture measure in 500 m DEM texture images were used as the best texture measures for classifying first-order landform in the areas of plain,hill and mountain.In the 30 m DEM texture images,due to the rich texture,the CNN pixel classification method,which is sensitive to the change of landform information,is utilized to participate in second-order landform classification,which greatly improved the accuracy of landform classification compared with other traditional methods.In this study,innovative research results have been obtained in optimizing parameters of texture calculation,constructing the DEM comprehensive texture factor,and accomplishing hierarchical landform classification.There is great theoretical value and application prospect in systematic analyzing the landform spatial features,classifying different landform types,discussing the evolution process of landform structure,clarifying the spatial pattern of landform morphology,revealing the formation and evolution mechanism of landform,and guiding the sustainable development of Shandong area.