关键词: Landsat 8影像, 蓝色顶面建筑物, 光谱模型, 提取

Abstract: As one of the important artificial target types in the basic geographic database, the extraction of blue roof buildings is of great significance to urban environmental monitoring, illegal building monitoring, urban planning and management. In view of the low accuracy and incomplete boundary of the existing building extraction methods, and the phenomenon of foreign matter with the same spectrum and different spectrum of the same object. Taking Luoyang as the research area, according to the principle of different spectral characteristics of different ground objects, a new spectral model which can enhance the blue top information and weaken the information of other non target ground objects is constructed in this paper. Firstly, the Landsat 8 image to be detected is radiometric calibrated and atmospheric corrected to obtain the surface reflection eigenvalue image after removing the atmospheric influence. Secondly, the samples of a certain number of pixels are collected for the blue features in the image and other interfering features that are difficult to distinguish from the blue features, and the spectral curve is generated. The band addition, difference and positive and negative processing are used to construct a new spectral index that can enhance the information of the blue features and weaken the information of other interfering features. Finally, the density segmentation method is used to obtain the most appropriate threshold range for extracting the blue top building, and then extract the blue top building in the image. The results showed that the overall classification accuracy of the new spectral index model is 94.46%, the Kappa coefficient is 0.889 2, the accuracy rate is 91.95%, and the recall rate is 95.19%, which are better than the maximum likelihood method. According to the extraction results, the area of the blue roof buildings in the urban area of Luoyang city is 16.28 km², accounting for 0.564% of the area of the study area and 0.107% of the total area of Luoyang city. The results show that this method has better extraction effect and accuracy than the maximum likelihood method, and has certain advantages and practicability.

Key words: Landsat 8 image, blue roof building, spectral model, extract