Authors

Abstract

Accurate spatial distribution of grassland degradation indicator species is of great significance for grassland degradation monitoring. In order to realize the intelligent remote sensing grassland degradation monitoring task, this paper collects remote sensing data of three degradation indicator species of desert grassland, namely, constructive species, dominant species, and companion species, through the UAV hyperspectral remote sensing platform, and proposes a multi-feature fusion (MFF) classification model. In addition, vertical convolution, horizontal convolution, and group convolution mechanisms are introduced to further reduce the number of model parameters and effectively improve the computational efficiency of the model. The results show that the overall accuracy and kappa coefficient of the model can reach 91.81% and 0.8473, respectively, and it also has better classification performance and computational efficiency compared to different deep learning classification models. This study provides a new method for high-precision and efficient fine classification study of degradation indicator species in grasslands.

Abstract in Chinese

准确掌握草地退化指示物种的空间分布对草地退化监测有着重要意义。为实现智能化遥感草地退化监测任务，本文通过无人机高光谱遥感平台对荒漠草原建群种、优势种和伴生种三种退化指示物种遥感数据采集，并提出一种多特征融合（MFF）的分类模型。此外，引入垂直卷积、水平卷积和分组卷积机制，进一步减少模型参数量，有效提升模型的计算效率。结果表明，该模型的总体精度和kappa系数分别可达91.81%、0.8473。同时，与不同深度学习分类模型相比，也具有更优的分类性能和计算效率。本研究为草地退化指示物种的高精度、高效率的精细分类提供了一种新方法。