丘陵区无人机影像基础信息提取方法初探

Preliminary study of basic information extraction method of UAV image in Hilly area

无人机影像具有高空间分辨率的特性,在体现地物细节上有着多光谱数据所不具有的优势,同时也使得常规的多光谱影像的信息提取方法并不适用于无人机影像,针对这一问题,面向对象法被应用于高分辨率影像信息提取中;与常规方法相比,面向对象法不仅仅是利用影像的光谱信息,也考虑了影像的纹理、颜色、形状指数等多种要素,大大地提高了高分辨率影像的信息提取精度。但面向对象法涉及要素多,需要对研究区有一定的了解,同时需要建立相应的分类规则。该文分别利用基于光谱分析的阈值法及基于对象的面向对象法,对居民区、丘陵区、水体及混合区进行基础地类信息提取,并从时间及精度2个角度进行对比分析。

With high spatial resolution characteristics, UAV images have advantages in reflecting the feature details which multi-spectral data do not have. Moreover, it makes the conventional information classification method of multispectral images do not apply to UAV images. Aiming at this problem, the object-oriented method was applied to extract the high resolution image information. Comparing to the con- ventional method, not only the spectral information of the object was used, but also the texture, color, shape index and other elements were taken account of, which greatly improved the accuracy of the infor- mation extraction of high-resolution images. However, object-oriented approach involves multiple ele- ments, which needs knowledge of the study area and establishing the appropriate classification rules. This paper extracted the class information of residential areas, hilly areas, water and mixed areas by using spectral analysis-based threshold method and object-based object-oriented method respectively. This paper analyzed the results from two angles： time and accuracy.