基于分块特征点密度的多特征融合遥感图像场景分类

Remote Sensing Image Classification Based on Fusion of Multiple Features with Block Feature Point Density Analysis

随着遥感等对地观测技术的发展,遥感图像分辨率越来越高。相比于中低分辨率遥感图像,高分辨率遥感图像能够提供更详细的地面信息,但各种地物空间结构分布较复杂。针对高分辨率遥感图像中的不同目标,各种特征有效性不尽相同、彼此存在互补现象,提出了一种分层多特征融合的场景分类方法。该方法首先对图像进行预分类,粗分为特征点分布均匀与不均匀两大类;然后,对分布均匀类别提取颜色直方图特征和Gabor纹理特征,对分布不均匀类别提取Sc SPM(基于稀疏编码的空间金字塔匹配)特征;最后分别训练支持向量机分类器对测试图像进行分类。在一个2100幅图像构成的大型遥感图像数据库上的实验结果表明,提出的算法比仅用单一特征分类方法的最高精度提高了10%;与其他融合方法相比,提出的方法取得了最高分类精度,达到了90.1%;算法时间复杂度也大为降低。

With the development of remote sensing and the related techniques, the resolution of these images is largely improved. Compared with moderate or low resolution images, high-resolution images can provide more detailed ground information. However, a variety of terrain has complex spatial distribution. The different objectives of high-resolution images have a variety of features. The effectiveness of these features is not the same, and some of them are complementary. Considering the above characteristics, a new method is proposed to classify remote sensing images based on the hierarchical fusion of multi-feature. Firstly, these images are pre-classified into two categories in terms of whether feature points are uniform or non-uniform distributed. Then, the color histogram and Gabor texture feature are extracted from the uniform distributed categories, and the Sc SPM（linear spatial pyramid matching using sparse coding） feature is obtained from the non-uniform distributed categories. Finally, the classification is performed by the two different support vector machine classifiers. The experimental results on a large remote sensing image database with 2100 images show that the overall classification accuracy is boosted by 10% in comparison with the highest accuracy of single feature. Compared with other methods of multiple features fusion, the proposed method has achieved the highest classification accuracy which has reached 90.1%, and the time complexity of the algorithm is also greatly reduced.