基于多波段像元形状指数点云和高光谱数据分类分析研究.pdfVIP

Geomatics Science and Technology 测绘科学技术, 2016, 4(4), 117-127 Published Online October 2016 in Hans. /journal/gst /10.12677/gst.2016.44014 A Multiple Bands Pixel Shape Index for Classification of LiDAR and Hyperspectral Data Bo Li State Key Laboratory of Remote Sensing and Information Engineering, Wuhan University, Wuhan Hubei th rd th Received: Oct. 14 , 2016; accepted: Oct. 23 , 2016; published: Oct. 26 , 2016 Copyright © 2016 by author and Hans Publishers Inc. This work is licensed under the Creative Commons Attribution International License (CC BY). /licenses/by/4.0/ Open Access Abstract In this paper, we proposed a spatial feature extraction method by fusing the hyperspectral image and LiDAR point cloud data. Conventional data fusion methods in classification are often the sim- ple combination of spectral and spatial information of the images and the height information of the LiDAR data. However, these methods cannot make full use of the complementary information of multi-source data. We proposed to extract the spatial features by fusing multi-source data. In order to obtain the shape index of high-dimensional fused data, we improved the pixel shape in- dex by applying the spectral angle distance measurement. The proposed method firstly fused the hyperspectral image and LiDAR data to enhance the heterogeneity between different classes. Then, the shape features are extracted with the SAD-based pixel shape index. Finally, the spectral, spa- tial and height information are prepared to the SVM classifier. The experiment shows that the proposed method achieves better results and accuracy, compared to the conventional means. Keywords Hyperspect