多源遥感数据融合的道路材质精细识别

针对当前道路识别方法对道路材质识别精度不高的问题,提出了一种融合多源遥感影像进行城市道路材质高精度识别的方法。首先,对高光谱遥感影像中的地物光谱曲线进行分析,在保留地物光谱分离度较大的波段的基础上提升计算效率;通过分步融合的策略对多源遥感影像进行融合,提升高光谱影像的空间分辨率,为后续道路材质识别提供高质量的数据保障。其次,通过使用不同的指数对融合影像进行掩膜,提取城市建筑物,并在此基础上提取建筑物纹理信息与光谱信息,进行多特征融合并分类;最后,通过影像后处理对提取的道路进一步进行规范,得到最终高质量的道路材质识别结果。通过使用高分五号高光谱影像、高分二号全色/多光谱影像、高分一号多光谱影像对提出的方法进行实验验证,试验结果表明,本文方法可取得较高精度的道路材质识别效果,具有较好的应用价值。

Fine Identification of Road Material Based on Multi-source Remote Sensing Data Fusion

In view of the problem that the current road recognition methods have low accuracy of road material recognition,in the paper,a method of fusing multi-source remote sensing images is proposed for high-precision recognition of urban road materials.Firstly,the spectral curve of ground objects in hyperspectral remote sensing images is analyzed,the bands with large spectral separation are reserved and the calculation efficiency is improved;the multi-source remote sensing images are fused step-by-step,which can improve the spatial resolution of hyperspectral images and provide high-quality data guarantee for subsequent road material identification.Secondly,the urban buildings are extracted by using different indices to mask the fused image;then,the texture information of the building area is extracted,combined with spectral information,the multi-feature fusion and classification are carried out;the extracted roads are further standardized by image post-processing,and the final high-quality road material identification results are obtained.The proposed method is verified by GF-5HS images,GF-2PAN/MS images,and GF-1MS images.The experimental results show that the step-to-step fusion strategy is suitable for the data fusion that has a big ratio of spatial resolution,the buildings can be extracted by multiple indices,and the proposed method can achieve good identification effect of road material,which proves the effectiveness of the method and its great application value.