Object-based binary encoding algorithm -an integration of hyperspectral data and DSM

Abstract

The advent of advanced processing techniques and high speed computers have led to the possibility of supplementary hyperspectral data with information about different kinds of object features that can be observed in the images, for example, shape and size. Other data sources, e.g., digital surface model from airborne laser scanning data, can provide height information for the object features. In this paper an improved binary encoding method (IBE) is proposed to integrate such additional information into the binary encoding matching method. The original binary encoding method proceeded spectral information pixel by pixel; IBE method is based on object-based classification. The hyperspectral and DSM data were corporately used in the method. During the method, the information of target objects was represented by 280 binary codes according to IBE rules, practical experiences and user requirements. We applied the proposed method to classify the test area. The results show that the proposed method needs less training data, lower computation cost and can gain higher classification accuracy. It is beneficial especially for limited spatial extent and great variation of the ground contents.