The Analysis Methods Of Pixel-based Scale Effect In Remote Sensing

Scale problem has been widely concerned by remote sensing community.people have not enough cognition on the connotation of scale in remote sensing,and the current remote sensing community has shortage of awareness to the scale-dependent mechanism,Since the remote sensing technology is a complex system.In addition,we couldn't effectively use different methods to study scale problem of the remote sensing data.That it isbecause there are many research methods on scale problems,most of them are independent of each other,and could not effectively compare the methods.Therefore,it is of great significance to establish a conceptual framework and a method framework for the scale-dependent of remote sensing.This paper analyzed the connotation of scale in remote sensing image,summarized the understanding to scale-dependent,and proposed improved fractal methods to analyze the scale-dependent of remote sensing image pixel observation based on in-depth analyzed the fractal mechanism of remote sensing image and the existing fractal model.In order to verify the reliability of the improved methods,the existing scale-dependent analysis method(information entropy method)was used to compare and analyze with these improved methods.The experimental data used Beijing QuickBird images,from which the 2 sub-regions of building and farmland were cut out as the research objects.Using MFBM,MDBM the 2 fractal models respectively calculated the fractal dimension of research data with the different observation scales,and using the information entropy model calculated the information entropy of corresponding data with different observation scales.Then calculated the correlation between the fractal dimension and information entropy.Analysis the experimental results show that both of them are all present the trend of first rise then descend with the increase of the spatial resolution,and in some of the same characteristic scales will present inflection points.Further analysis shows that these characteristic scales corresponding to the inflection points and the sizes of image features have a certain relationship,which making the mixed pixels of the image are less.Therefore,these inflection points have some indicative significance for observing this feature.In order to systematically analyze the research method of scale problem of remote sensing image,this paper in-depth studies the two commonly used methods of statistical problem based on statistics(Improved Local Variance Methodand Semi-variogramMethod)from the theory to the method.And the above-mentioned experimental data are used to carry out a series of experimental experiments and analysis.Based on the above theoretical analysis of different statistical and fractal methods,the application ranges of different methods are summarized,and the experimental results of the above methods are compared and analyzed.Through this study,it is known that using the improved fractal methods research the pixel scale-dependent of remote sensing image,and breaking the category of previous observation scale method to analyze the observation scale problem of remote sensing image,from different angles,has a certain theoretical and guiding significance for GIS research and geography application.In addition,by the comprehensive analysis of the theoretical and experimental results of different methods,the scopes of application of different methods are summed up,which has important theoretical and guiding significance to the selection of research methods in future study of scale.