Answers are grounded in this paper and cite their sources.
Leaf Area Index (LAI) is a key parameter for assessing vegetation canopy structure, function, and growth dynamics, playing a pivotal role in climate change and ecosystem studies. Hemispherical photography is a widely used technique for rapid LAI estimation; however, its accuracy is often affected by the structural complexity of vegetation canopies and the inherent simplifications in estimation models, resulting in persistent uncertainties in the obtained estimation. To address these limitations, this study leverages deep learning to enhance LAI estimation based on hemispherical canopy images. The main objectives and results are as follows: (1) Seventeen simulated scenes representing diverse typical vegetation types were constructed using 3D modelling software to generate temporally extensive hemispherical image sequences from multiple sampling points. These simulated images were combined with in-situ field observations, including hemispherical photographs and corresponding LAI estimation, to form a comprehensive dataset comprising 12,916 samples . (2) An end-to-end deep regression model was proposed to estimate LAI directly from hemispherical images, with a systematic evaluation of different feature extraction networks, regression layers, and loss functions. (3) Comparative regression analyses based on both simulated data and LAI-2200 reference estimation demonstrated that the proposed approach achieved coefficient of determination ( R 2 ) values exceeding 0.91 . Compared to non-end-to-end approaches, it increased R 2 by 0.26–0.39 and simultaneously reduced R M S E by 0.82–1.30 and M A E by 0.64–1.01. These results indicate that the proposed end-to-end deep learning-based framework can substantially improve LAI estimation performance from hemispherical images under the current reference and evaluation settings.
No takes yet. Share an insight, caveat, or question.
Deng et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: