A Novel Deep Learning Network for Fractional Vegetation Cover Calculation Using Full-Waveform LiDAR

Traditional fractional vegetation cover (FVC) calculation methods relying solely on either airborne LiDAR waveforms or point clouds often suffer from degraded performance due to single-modality limitations and spatially uneven point density. To address this issue, this study proposes a multimodal fusion network termed the joint LiDAR waveform and point-cloud network (JWP-Net) and a 2D grid area ratio-based FVC calculation method. First, with a high-resolution digital orthophoto map as an auxiliary reference, an interactive human–machine visual interpretation workflow is adopted to precisely label LiDAR echoes into three categories: vegetation, land, and water. A standardized dataset is thereby established for subsequent data analysis and model training. Second, a novel deep learning network termed JWP-Net, which utilizes LiDAR echo waveforms and point clouds as multimodal inputs, is proposed to classify the laser points into vegetation, land, and water. Third, a 2D grid area ratio-based FVC calculation method is proposed to calculate the FVC based on the classified laser points. To validate the effectiveness and superiority of the proposed framework, experiments are conducted using raw measurement data acquired by the Optech Aquarius LiDAR system. Experimental results demonstrate that JWP-Net achieves an overall classification accuracy of 94.00% ± 0.41% (mean value ± standard deviation of 5-fold overall accuracies evaluated on the untouched held-out test grids), outperforming single-modal models by 1.77–2.52%. Furthermore, integrating DBSCAN-based spatial morphological smoothing on the held-out grids steadily improves the overall classification accuracy to 94.81% ± 0.44%. Moreover, compared with the traditional laser point-count-based FVC calculation method, the proposed 2D grid area ratio-based method reduces the FVC absolute deviation by 3 percentage points on test grids at a 0.2 m grid resolution. The multimodal fusion strategy that combines airborne laser full-waveform and point-cloud data, and the 2D grid area-ratio-based FVC calculation method proposed in this work, offer a novel research approach for high-precision FVC calculation using full-waveform LiDAR.

Authors

Institutions

Publication Details

Journal
Remote Sensing
Published
2026-11-16
DOI
https://doi.org/10.3390/rs18223928
Primary Topic
Remote Sensing and LiDAR Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A Novel Deep Learning Network for Fractional Vegetation Cover Calculation Using Full-Waveform LiDAR

Pingping Liu, Kanglin Cong, Xinglei Zhao, Mengyan Xu et al.
1 links
Remote Sensing
Remote Sensing and LiDAR Applications
article

A Novel Deep Learning Network for Fractional Vegetation Cover Calculation Using Full-Waveform LiDAR

Pingping Liu, Kanglin Cong, Xinglei Zhao, Mengyan Xu, Yanhua Zhu, Xiaoliang Han, Zhen Chen, Junru Zhao
article en

Abstract

Traditional fractional vegetation cover (FVC) calculation methods relying solely on either airborne LiDAR waveforms or point clouds often suffer from degraded performance due to single-modality limitations and spatially uneven point density. To address this issue, this study proposes a multimodal fusion network termed the joint LiDAR waveform and point-cloud network (JWP-Net) and a 2D grid area ratio-based FVC calculation method. First, with a high-resolution digital orthophoto map as an auxiliary reference, an interactive human–machine visual interpretation workflow is adopted to precisely label LiDAR echoes into three categories: vegetation, land, and water. A standardized dataset is thereby established for subsequent data analysis and model training. Second, a novel deep learning network termed JWP-Net, which utilizes LiDAR echo waveforms and point clouds as multimodal inputs, is proposed to classify the laser points into vegetation, land, and water. Third, a 2D grid area ratio-based FVC calculation method is proposed to calculate the FVC based on the classified laser points. To validate the effectiveness and superiority of the proposed framework, experiments are conducted using raw measurement data acquired by the Optech Aquarius LiDAR system. Experimental results demonstrate that JWP-Net achieves an overall classification accuracy of 94.00% ± 0.41% (mean value ± standard deviation of 5-fold overall accuracies evaluated on the untouched held-out test grids), outperforming single-modal models by 1.77–2.52%. Furthermore, integrating DBSCAN-based spatial morphological smoothing on the held-out grids steadily improves the overall classification accuracy to 94.81% ± 0.44%. Moreover, compared with the traditional laser point-count-based FVC calculation method, the proposed 2D grid area ratio-based method reduces the FVC absolute deviation by 3 percentage points on test grids at a 0.2 m grid resolution. The multimodal fusion strategy that combines airborne laser full-waveform and point-cloud data, and the 2D grid area-ratio-based FVC calculation method proposed in this work, offer a novel research approach for high-precision FVC calculation using full-waveform LiDAR.

Remote SensingVol. 18(22)
Shandong Agricultural University (CN)
Remote Sensing and LiDAR Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon