摘要
植被高度是评估森林资源数量和质量的重要参数之一,极化干涉合成孔径雷达(Polarimetric Interferometry SAR,PolInSAR)技术在单基线情况下,容易出现低矮树木高估现象。相比之下,激光雷达(Light Detection and Ranging,LiDAR)数据能提供更高的垂直精度,但其受到成本和算力的限制只适用于小规模场景。为了克服二者缺点,提出一种基于Transformer网络联合PolInSAR和LiDAR数据进行数据融合的网络模型——Poliformer网络。网络采用了卷积操作来合并来自两种不同源的数据,通过在生成的产品基础上进行数据优化和逼近处理,显著提升了最终输出的精确度,实现了更为准确、全面的地表高度和植被结构信息。
Vegetation height is a key parameter for evaluating the quantity and quality of forest resources.Polarimetric Interferometry Synthetic Aperture Radar(PolInSAR)technology,particularly in single-baseline scenarios,often leads to the overestimation of the height of low-lying vegetation.In contrast,Light Detection and Ranging(Li-DAR)data offer higher vertical accuracy but are constrained by cost and computational power,limiting their use to small-scale scenarios.To address these limitations,this paper introduces a novel network model,named Poliformer,which fuses data from both PolInSAR and LiDAR using a Transformer-based network.The network employs convolutional operations to merge data from these two diverse sources.By optimizing and approximating data basedon the generated products,the Poliformer significantly enhances the precision of the final output.This approach leads to more accurate and comprehensive surface height and vegetation structure information.
作者
李同
刘丽萍
孙学宏
陈劲烨
LI Tong;LIU Li-ping;SUN Xue-hong;CHENG Jin−ye(School of Electronic and Electrical Engineering,Ningxia University,Ningxia Yinchuan 750021;Ningxia Key Laboratory of Intelligent Perception for Desert Information,Ningxia Yinchuan 750021,China)
出处
《计算机仿真》
2025年第11期309-314,共6页
Computer Simulation
基金
星载L/P双波段Po-InSAR植被垂直结构模型与反演方法研究(62061038)。
