I. Introduction
Land-cover classification plays an important role in a variety of applications, such as Earth observation, environmental science, and forest management, etc. [1], [2]. Meanwhile, the development of sensor technology has provided varied data sources support for classification tasks [3]–[5]. Among these data, hyperspectral image (HSI) contains hundreds of spectral bands, which provides detailed spectral information of land covers [6]–[8]. However, its passive imaging mode makes it susceptible to cloudy weather, and difficult to discriminate objects with similar spectral reflectance. Conversely, the active acquisition of light detection and ranging (LiDAR) data is less sensitive to the weather condition. Specifically, LiDAR data are able to capture elevation information about surveyed area, which can help assess the size and shape of specific objects [9]–[11]. Currently, there is an increasing interest in collaborative utilization of HSI and LiDAR data for accurate land-cover classification, and diversified mutual-aid modes are investigated to provide comprehensive interpretation for the study area [12]–[14].