Research Of Impervious Surface Extraction Algorithm For High Resolution Remote Sensing Image

The rapid development of the city has led to the expansion of buildings with poor water permeability,such as buildings and roads.The original vegetation,land,and farmland on the ground surface have been replaced by objects with good water permeability.The increase in urban imperviousness has led to a series of social effects,such as water quality deterioration,urban heat island effects and other ecological and environmental problems.Accurate extraction of impervious surfaces provides guidance for many areas such as urban planning,heat island effect analysis,and sustainable city development.In traditional remote sensing image analysis,the main basis for classification is the broad-spectrum features of the features.In high-resolution remote sensing images,the amount of spectral information is particularly rich,and microscopic texture information is also very detailed.Therefore,the use of high-resolution remote sensing images to study impervious surfaces is a quick and effective technique.High-resolution remote sensing images cover a wide range of fields and have great advantages and potentials in the extraction of land objects.They have become the main technical means for obtaining high-precision urban impervious surfaces.This paper uses the Yanshan campus of Guangxi Normal University as the research area to extract the impervious surface of the city.The high-resolution remote sensing image is used to extract the impervious surface of the area.The classification methods of quantum genetic maximum entropy multi-threshold segmentation and multi-feature fusion are mainly described.The calculation methods of color information and texture information in high-resolution images are explained in detail.The experimental results are compared and analyzed.The Python programming language is used.A set of simple image processing software was designed.Finally,the work of the paper's research was summarized and forecasted.The main work of the dissertation is:(1)In order to solve the traditional threshold segmentation method can't meet the segmentation requirements of complex urban features,and produce serious "over-segmentation","under-segmentation" problem,this paper first uses a quantum-based genetic maximum threshold entropy segmentation method.Based on the analysis and study of the principle of maximum entropy partition,combined with the characteristics of population diversity and fast convergence rate of quantum genetic algorithm,quantum genetic algorithm is applied to multi-threshold segmentation of two-dimensional maximum entropy image threshold segmentation as an image,not only Effectively reduce the complexity of two-dimensional maximum entropy image segmentation calculation,and compared with the traditional threshold segmentation,the impervious surface extraction accuracy is improved,and the operation time is shortened.(2)In order to obtain the training samples needed in the experiment,the original image was marked by a manual labeling method,and the pervious surface and the impervious surface were marked with different colors,and some representative areas were marked as far as possible to obtain the final result.Experimental training samples.(3)In order to make full use of the rich color information and texture information in highresolution images,an impervious surface extraction method based on multi-feature fusion is proposed.This method firstly uses the color information in the image to color the color itself.The RGB features are combined with the characteristics of the color space,normalized into a feature vector,and then using the texture information in the image,different size windows are used to slide-out the texture features.Using these two features,the SVM classifier and the BP neural network are used respectively.The whole image is classified according to pixels.The results show that the impervious surface extraction accuracy of this method is higher than the traditional method.(4)Using the Python programming language,call OpenCV image processing library functions and PyQt interface library functions.The impervious surface extraction system is designed for the method used in this paper,and the calculation results and calculation time of different methods are compared.