Home > Published Issues > 2016 > Volume 7, No. 4, November 2016 >

Tensor Locality Preserving Projections Based Urban Building Areas Extraction from High-Resolution SAR Images

Bo Cheng, Shiai Cui, and Ting Li
Institute of Remote Sensing and Digital Earth Chinese Academy of Sciences, China

Abstract—Currently, the majority of Manifold Learning algorithms applied for SAR image feature extraction are vector based; the For tensor based SAR images, a “convert to vector: process has to be taken before attribute extraction. During this process, curse of dimensionality would be occurred and information of space geometry structure could be lost. Those phenomenon are not conducive for target recognition of SAR images. In this paper, Radarsat-2 images were used as experimental data and the Tensor Locality Preserving Projections (TLPP) algorithm was applied for the attribute extraction of high-resolution SAR images, to improve the recognition accuracy and achieve fast extraction of urban building areas. A comparison was made for the recognition results of TLPP and Locality Preserving Projections (LPP). It is found that that TLPP algorithm has a strong adaptability of generalization, which indicates that TLPP can be effectively used for fast extraction of urban building areas from high-resolution SAR images, with high accuracy.

Index Terms—synthetic aperture radar, manifold learning, feature extraction, tensor locality preserving projections

Cite: Bo Cheng, Shiai Cui, and Ting Li, "Tensor Locality Preserving Projections Based Urban Building Areas Extraction from High-Resolution SAR Images," Vol. 7, No. 4, pp. 291-296, November, 2016. doi: 10.12720/jait.7.4.291-296