Study On Land Cover Classification Method Based On OSM Tags

Research on land use and land cover(LULC)is one of the most important ways for the study of global change.LULC are driving factors impacting global climate.Since Land Use and Cover Change(LUCC)project was launched by International Geosphere-Biosphere Programme(IGBP)and International Human Dimensions Programme on Global Environmental Change(IHDP)in 1995,how to acquire up-to-date and accurate land cover data has become a research host spot among geoscience scholars.Remote sensing technology is the most important mean for obtaining land cover data.At present,remote sensing automatic monitoring of land cover in the world is mainly based on low-resolution remote sensing data(300m-1km),and besides its' low spatial resolution,these products usually have long time span,and large system differences.Although mediumresolution remote sensing data has been used to monitor land cover,there are also problems such as relying on visual interpretation or human interaction,long production cycle,and insufficient automation.The automatic production and dynamic updating of high-resolution land cover products is one of the hot and difficult issues in remote sensing research.In recent years,along with the rapid development of Volunteered Geographic Information(VGI),WebGIS and electronic maps,VGI have the potential to open new avenues for LULC mapping.The OpenStreetMap(OSM)project is an ongoing effort which generated spatial and thematic content on a planetary scale since 2004 through millions of volunteer contributors within an open source environment.Original OSM data stored in vector structure is compact and has low redundancy,which is convenient for describing the boundary of geographic features.The tags contained in the OSM data provides the possibility of using OSM data to assist remote sensing images for land cover classification.To sum up,this paper studies the automatic and fast acquisition method for land cover classification based on OSM tags with OSM protocolbuffer binary file(*.pbf)and WorldView-2 high spatial resolution image in Washington D.C.The main research contents and research results of the thesis are as follows:(1)The relationship model between OSM tag type and land cover classification system.Referring to the existing LULC classification system,the relationship model between OSM tags and land cover types in the study area was completed based on the OSM tag,and the land cover classification system based on OSM tags was established.(2)Based on the research(1),a method for acquiring dynamic land cover data based on OSM is proposed.The basic procedures are as follows: firstly,extract land cover results in research area based on OSM tag data;secondly,in case the study area is small,the land cover in the study area is supplemented regarding to various electronic maps;in case the study area is large,the OSM land cover classification results are used as training samples for supervised remote sensing image classification,finally the complete land cover classification result is obtained.(3)Evaluation of OSM land cover classification accuracy in the study area.Based on the classification method proposed in(2),the land cover map in Washington D.C.is carried out,and the classification accuracy is qualitatively evaluated by comparing with electronic maps.Combined with high-resolution remote sensing image,the maximum likelihood classification is performed and the confusion matrix is calculated to quantitatively evaluate the classification accuracy.Compared with the classification results of other land cover classification data,the proposed method has better classification efficiency and higher classification consistency in this study area.