Relative Total Variation Structure Analysis-Based Fusion Method for Hyperspectral and LiDAR Data Classification

Yinghui Quan,Mengdao Xing,Wenjiang Huang,Gabriel Dauphin,Wei Feng,Yingping Tong,Wentao Zhu

doi:10.3390/rs13061143

Abstract

The fusion of the hyperspectral image (HSI) and the light detecting and ranging (LiDAR) data has a wide range of applications. This paper proposes a novel feature fusion method for urban area classification, namely the relative total variation structure analysis (RTVSA), to combine various features derived from HSI and LiDAR data. In the feature extraction stage, a variety of high-performance methods including the extended multi-attribute profile, Gabor filter, and local binary pattern are used to extract the features of the input data. The relative total variation is then applied to remove useless texture information of the processed data. Finally, nonparametric weighted feature extraction is adopted to reduce the dimensions. Random forest and convolutional neural networks are utilized to evaluate the fusion images. Experiments conducted on two urban Houston University datasets (including Houston 2012 and the training portion of Houston 2017) demonstrate that the proposed method can extract the structural correlation from heterogeneous data, withstand a noise well, and improve the land cover classification accuracy.

Highlights

The advancement of remote sensing technologies has resulted in an improvement in the availability of multi-sensor data in the same region and a deeper understanding of the research area [1,2]
For the 2012 Houston dataset, the relative total variation structure analysis (RTVSA) fusion method was applied based on the local binary pattern (LBP), extended multi-attribute profile (EMAP), and Gabor feature extraction, respectively
The spatial information extracted by the EMAP can significantly improve the classification accuracy

Summary

Introduction

The advancement of remote sensing technologies has resulted in an improvement in the availability of multi-sensor data in the same region and a deeper understanding of the research area [1,2]. The hyperspectral image (HSI) has hundreds of spectral bands for each pixel. HSI may not be able to reliably identify objects with the same spectral properties [4]. The light detecting and ranging (LiDAR) data could provide height information that is complementary to spectral details [1,5]. Objects of the same elevation but made from different materials cannot be separated using only LiDAR elevation information [6]. The fusion of the high spectral resolution of HSI and the structural information given by LiDAR will provide more complete and enhanced surface properties for a broader range of applications, such as forest monitoring [7,8,9], biomass estimation [10], and geological analysis [11]

Objectives

Results

Discussion

Conclusion