Research On Urban Environment Structure Information Extraction Based On High-Resolution Remote Sensing Image

The development of the remote sensing technology made us obtain abundant information, especially with high spatial resoluteion QuickBird remote sensing image, which could provide important information resources for urban envirnomnent field. Among many urban spatiall information, environment structure information was one of the most important, it was the basis of urban environment quality evaluation, urban planning.This paper mainly researched preprocessing and classification of QuickBird image, in combination with relief map information for the examination of a test area with representative environmental structure in Dandong city. In details, the study in this paper mainly included the following aspects:1. In image preprocessing, due to the flat terrain of the study area, a polynomial transform method was adopted to solve pixel position deviation using collected ground control points with obvious features through GPS. Furthermore, to remove the effect of the building shadows in high spatial resolution image on the spectra of land cover objects, the building shadows are automatically extracted by eCognition software and corrected by the Lambertian model.2. According to the characteristics of various land cover classes and their different influence on the urban environment, we established a classification system of urban environmental structure, which has four classes: construction space, transportation space, green space and water space. Based on these, the space can be divided into thirteen sub-classes, i.e., large building, small building, bare land, main road, sidewalk, port, park, arbor, shrub, artificial greenbelt, yield greenbelt, river and lake.3. In this paper, pixel based classification and object-oriented classification two kinds of classification methods were applied to classify the high resolution remote sensing image of a region in Dandong city with typical urban environmental structure feature, with an accuracy assessment using 300 points generated randomly by the computer. The result showed that the object-oriented classification method was more appropriate for high spatial resolution remote sensing image, because QuickBird image has obvious features such as shape, texture, et al., which can be used to extract important information, although, at low spectral resolution.The innovate of the paper was using high resolution remote sensing image though precise geometric correction, image blend, shadow correction, urban environmental structure was researched through introduced object-oriented classification, indicated influence of urban development to the environment structure through compared with aerial photograph.We drew some conclusions that, in urban environmental structure research, automatic classification technique of high spatial resolution remote sensing image can replace traditionally visual interpretation method on aerial photograph, and acquire 86.67% classification accuracy. The comparison of pixel based classification and object-oriented classification indicated the object-oriented method, utilizing sufficiently information, such as shape, texture, and area, et al. in high spatial resolution image, and providing classification results fast and accurately, was proven an efficient classification method with broad development potential, this has important theories meaning and applied value to the urban environment structure research.