Object Oriented Classification Technology For Landscape Information Extraction

Classification in pixel-based image processing techniques can not be satisfactory for high resolution image, because conventional classification algorithm attempts to classify each pixel based on this spectral information of the image mainly. The high resolution images only have three or four bands, especially the merging true-color images which only have three bands on the visible light range, and their spectral information is weak relatively, in other words, having no enough basis. Actually, the high resolution images, such as Spot5, Ikonos and Quickbird data, have mostly abundant shape, structure and texture information. The object-oriented image classification concept which is concerted recent years uses not only spectral information, and also spacial structure and texture information, even experience and knowledge of the human visual recognition. Ecognition follows the concept that important semantic information necessary to interpret an image is not represented in single pixels but in meaningful image objects and their mutual relations. Therefore, the software delivers results noticeably better than conventional methods. It leads to higher classification accuracy.This paper discusses eCognition, a professional object oriented remote sensing classification software developed by Definiens Imaging GmbH in Germany, and introduces object oriented classification concept and technique into landscape information obtaining field, which has realized information retrieving of Landscape ecological and landscape morphology, through multi-scale image division, establish of class hierarchy structure and semantic structure, classification role and reference.The research in this paper is based on kernel scene in ZhangJiajie, which uses 2.5 meter spatial resolution of SPOT5 merging images, and has successfully obtained 23 types of landscape ecological information including forest vegetation and many types of Land utilization; This paper uses fine extraction of "peak forests and stelae " shadow to get some information of stele in scene such as number, height and their distribution, which provides scientific evidence for the Ancient legends of "marvelous spectacle with three thousands of peak forests" and new Technology for the research of landscape ecological and landscape morphology.