Research On Object-oriented Classification For Remote Sensing Based On Superpixel

With the development of remote sensing imaging technology and the improvement of multi-source image’s resolution in satellite visible light, multi-spectral and hyper spectral, the high resolution remote sensing image has been widely used in various fields. The traditional scene classification method based on pixel is lack of accuracy to meet the requirement of high resolution remote sensing image processing, because of its inability of extracting the same object type’s structure and texture features. So the object-oriented remote sensing image classification method is proposed to solve the accuracy problem.Object-oriented remote sensing image classification include object segmentation and object classification. It presents the high resolution remote sensing image segmentation technology, and discusses the super pixel segmentation algorithm based on SLIC. This algorithm produces objects with strong homogeneity and compact structure, and extracts features of these objects that clearly reflect the typical structural information and spatial topological relationship. Thus the algorithm provides efficient and reliable segmentation strategy.This paper also presents an object-oriented remote sensing image scene classification method based on super pixel segmentation. The method is consist of typical objects classification generation, super pixel segmentation based on SLIC, texture and spectral features extraction of typical objects and supervising classification based on supporting vector machine. Remote sensing image classification software system is designed and implemented to meet the requirements.This paper discussed the influence of feature extraction, cluster and the scale of segmentation.The classification experiments on some visible light and infrared image show that the method has good applicability in high resolution remote sensing image scene classification.