Related Experiment Video
Updated: Aug 5, 2026
08:47
Computer Vision-Based Biomass Estimation for Invasive Plants
Published on: February 9, 2024
Improving aboveground biomass estimation in desert steppe using hyperspectral semantic segmentation
Xiaotian Sun1, Haichao Wang1, Zhiyong Pei1
1College of Materials Science and Art Design, Inner Mongolia Agricultural University, Hohhot, China.
Frontiers in Plant Science
|July 30, 2026
Summary
Accurate aboveground biomass (AGB) estimation in desert steppes is improved using hyperspectral semantic segmentation to enhance vegetation spectral purity. This method effectively reduces background contamination for better ecosystem productivity monitoring.
Area of Science:
- Remote Sensing
- Ecology
- Biomass Estimation
Background:
- Accurate aboveground biomass (AGB) estimation is crucial for desert steppe ecosystem assessment.
- Sparse vegetation and soil backgrounds complicate AGB retrieval, reducing model accuracy.
Purpose of the Study:
- To develop a hyperspectral framework for improved AGB estimation in desert steppes.
- To enhance vegetation spectral purity and model performance via semantic segmentation and spectral feature optimization.
Main Methods:
- Utilized hyperspectral imagery from Inner Mongolia, China.
- Applied semantic segmentation (U-Net, SegNet, DeepLabV3+, RF) for vegetation extraction.
- Optimized spectra (preprocessing, VI, feature selection) followed by PLSR for AGB estimation.
Main Results:
- U-Net achieved the highest vegetation extraction accuracy (OA=0.923).
- The optimal model (SNV+VI+CARS+PLSR) showed high performance (R²=0.83, RMSE=18.93 g/m², RPD=2.39).
Conclusions:
- Hyperspectral semantic segmentation significantly improves AGB estimation by reducing background noise.
- The proposed framework offers an effective solution for biomass estimation in sparse grasslands.