基于无人机高光谱的滨海湿地智能制图方法探讨

Exploration of Intelligent Mapping Method for Coastal Wetlands Based on Unmanned Aerial Vehicle Hyperspectral Data

滨海湿地遥感智能制图可为政府管理部门及时掌握生态环境变化、生态修复成效提供数据支持。本文以无人机高光谱为数据源,构建了一种面向有限样本的轻量化多尺度改进3DCNN模型,在减少计算量的同时高效获取高光谱数据的空间特征和全局光谱特征,实现滨海湿地典型区域的智能解译制图。结果表明,本文设计的模型在参数量远小于其他方法的前提下,获得了更好制图精度,为滨海湿地快速、精细遥感制图提供了参考。

Remote sensing intelligent mapping of coastal wetlands can provide data support for government management departments to timely grasp changes in the ecological environment and the effectiveness of ecological restoration.In this paper,we used unmanned aerial vehicle hyperspectral data source to construct a lightweight multi-scale improved 3D CNN model under limited samples.The model reduced computational costs while efficiently obtains spatial and global spectral features of hyperspectral data,and achieved intelligent interpretation and mapping of typical coastal wetland areas.The experimental results showed that the proposed model achieved better mapping accuracy,with much smaller parameter count than other methods,providing a reference for quick and precise remote sensing mapping of coastal wetlands.