Research On Impervious Surface Information Extraction And Dynamic Change Of Beijing Based On Landsat And Sentinel-1 Data

The acceleration of urbanization has prompted the rapid expansion of impervious surfaces.A large number of impervious surfaces have replaced the natural landscape dominated by green vegetation,which has had a serious negative impact on the urban ecological environment.As a microcosm of the rise of Chinese cities,Beijing's urbanization process is more rapid.A series of environmental problems such as the urban heat island effect,deterioration of water quality and urban waterlogging have become huge challenges for Beijing's urban development.Timely and accurate access to impervious surface information is of great significance to guide the planning of the capital city and protect the urban ecological environment.At present,many remote sensing inversion methods for impervious surface have achieved high accuracy,but the confusion between impervious surfaces and bare soil,sand and other objects is still the main factor that plagues the accurate extraction of impervious surfaces.With the development of remote sensing technology,multi-source remote sensing data are constantly emerging and fusion of multi-source data can obtain richer and more accurate information than single data.Therefore,using the advantages of remote sensing data from different platforms to complement each other can be used as an important way to improve the accuracy of impervious surface extraction.In this paper,the applicability of integrated Landsat multispectral data and Sentinel-1 imaging radar data to extract impervious surface is explored in the study area within the Beijing Sixth Ring Road.Firstly,the optimal algorithm for extracting impervious surface information is analyzed.That is,based on the Landsat multispectral band,the maximum likelihood method,support vector machine,artificial neural network and random forest algorithm are used to classify the image to extract the impervious surface information.Experiments show that the random forest algorithm can obtain high impervious surface extraction accuracy.And then,three groups of experiments were designed in the study,random forest algorithm is used to extract impervious surface information based on LT_OPT model using only Landsat image spectral information,LT_OPT+LT_TXT model combining Landsat image spectral information and texture information,LT_OPT+S1_TXT model integrating Landsat image spectral information and Sentinel-1 image texture information,respectively.Experiments show that the overall accuracy of the classification using the LT_OPT+S1_TXT model is 97%,and the Kappa coefficient reaches 0.963.Particularly,the extraction accuracy of impervious surface reaches 97.08%.Compared with the LT_OPT model and the LT_OPT+LT_TXT model,the classification accuracy is improved by 11.25% and 6.25%,respectively.In addition,the geometric shape and contour of the features on the classified image based on the LT_OPT+S1_TXT model are more obvious.It is proved that the combination of Landsat and Sentinel-1 data can effectively reduce the confusion between impervious surface and bare soil,obtain high classification accuracy of impervious surface and can inhibit the phenomenon of “salt and pepper” often associated with image classification.Compared with the texture information of Landsat multispectral image,Sentinel-1 radar image has more abundant texture information,which can significantly improve the ability to distinguish objects.Since the Sentinel-1 data was gradually applied to application in October 2014,therefore,the Landsat data of 1996,2003,2010 and 2017 were used to obtain the spatial distribution of the impervious surface in the study area with the LT_OPT+LT_TXT model.Based on this,the dynamic characteristics of the impervious surface in Beijing were analyzed.The results show that the impervious surface change between the fifth ring and the sixth ring is most significant between 1996 and 2017,while the variation of the impervious surface between other loops is small.From 1996 to 2003,impervious surface in the study area expanded rapidly towards the north,south and east part,with an increase of 17.61%.Notably,the impervious surface growth between the fifth ring and the sixth ring reaches 15.29%.From 2003 to 2010,the impervious surface between the fifth ring and sixth ring maintained a small increase,and the impervious surface between other loops decreased.Up to 2010,the impervious surface inner the Beijing sixth ring Road accounted for 70.98% and reached the maximum.In 2010-2017,the impervious surface between the fifth ring and sixth ring was significantly reduced by 5.65%,while the proportion of impervious surfaces between other loops continued to decline.