Technology Research Of Land Cover Change Detection In Coastal Areas Based On Object-oriented Classification

Land cover change detection is not only the essential part of the geographic conditions monitoring,but also the important content of the innovation and development of surveying and mapping geographic information.As a part zone of the national general survey of geographic conditions,the coastal area is one of the important research areas of land cover change detection.With the development of remote sensing technology,the high-resolution remote sensing image,vector data and the technology of land cover classification provide important data sources and methods for land cover change detection in coastal areas.However,due to the characteristics of coastal areas,such as special geographical location,advanced economy,rapid change and patch fragmentation,how to realize the land cover change detection in coastal areas has become a current hotspot issue.In this study,in view of the combination of singe-temporal high-resolution remote sensing image and vector data,land cover change detection in coastal areas is developed based on the object-oriented classification with optimal feature variables set.This paper established an appropriate classification system of land cover according to the characteristics of coastal areas.And the optimal feature variables set which is fit to the object-oriented Global Land Cover(GLC)classification method.Then coastlines will be extracted by using the object-oriented GLC classification method with optimal feature variables set.Additionally,to complete the change detection based on the combination of singe-temporal high resolution remote sensing image and vector data,the multi-scale segmentation for post-temporal image was performed with the guidance of the old land cover vector data,and the object-oriented GLC classification method with optimal feature variables set was utilized for post-temporal image classification.The main tasks and contributions are as follows:(1)We firstly established an appropriate classification system of land cover for coastal areas and produced the object-oriented GLC classification method with optimal feature variables set based on the RelifF algorithm and PCA algorithm and the accuracy of the object-oriented GLC classification as the evaluation criterion.It is proved that the method with optimal feature variables set selection can improve the classification accuracy of object-oriented GLC,and the result shows that the overall accuracy and Kappa coefficient of the object-oriented GLC classification with optimal feature variables set separately increase by 14.7688%and 0.1450 over the overall accuracy and Kappa coefficient of the object-oriented GLC classification without optimal feature variables set.Besides,experiments results indicated that the object-oriented GLC classification method with optimal feature variables set outperforms SVM and ANN with obvious highest overall accuracy(90.0609%)and Kappa coefficient(0.8792).(2)We extract the coastline using the object-oriented GLC classification method with optimal feature variables set,which developed in this paper.Performances of the object-oriented GLC classification method with optimal feature variables set are qualitatively evaluated and quantitatively evaluated through experiments.The results showed that the method developed in this paper has a better performance to extract coastlines than SVM and a modified Canny edge detection algorithom.(3)This paper determine the study scope of land cover change detection in coastal areas after coastline extraction,and in view of the combination of singe-temporal high-resolution remote sensing image and vector data,we use vector data to guide the multi-scale segmentation for post-temporal image.Moreover,the object-oriented GLC classification method with optimal feature variables set is applied to the classification for post-temporal image to improve the accuracy of change detection.(4)In this paper,a coastal area in She Yang of JiangSu province was taken as the experimentation area.By using the high-resolution remote sensing image of Resources Satellite Three(ZY-3)in 2015 and the land cover vector data in 2012,the experiment was carried out to verify the feasibility and effectiveness of the method and process in above(3)proposed in this paper.And the process and framework of land cover change detection in coastal areas were finally developed based on the the results of(2)?(3)and(4).