Abstract: Leaf area index (LAI) is an important biophysical parameter for characterizing crop canopy structure and growth status, and its accurate estimation is essential for high-throughput phenotyping and precision breeding at the breeding plot scale. With the rapid development of unmanned aerial vehicle (UAV) remote sensing technology, UAV multispectral imagery has become an efficient tool for monitoring crop growth. However, most existing LAI inversion methods mainly rely on empirical statistical relationships or one-dimensional radiative transfer models, which often overlook the complex three-dimensional canopy structure of row-planted crops. Consequently, the estimation accuracy and stability of these methods are often affected by canopy heterogeneity, growth stages, and varietal differences, particularly under breeding conditions with complex canopy architectures.To address these limitations, this study developed a winter wheat LAI inversion method based on a three-dimensional radiative transfer model (3D RTM) and UAV multispectral observations. Based on the Digital Plant Phenotyping Platform (D3P), a three-dimensional wheat canopy simulation system was constructed to represent canopy structural variability under different growth conditions. Using the 3D RTM, 10,000 groups of simulated canopy reflectance datasets were generated and used to establish a physically constrained inversion dataset. Compared with traditional empirical approaches, the method explicitly incorporated canopy structural information and radiative transfer constraints, thereby improving the robustness and transferability of LAI inversion under varying observation conditions.To evaluate the performance of the method, multi-temporal UAV multispectral observations and synchronous field LAI measurements were collected throughout the entire growing season at two representative experimental sites, including Baima in Nanjing and Yangling in Shaanxi Province. The datasets covered multiple wheat varieties and different growth stages from jointing to grain filling, resulting in more than 2,500 validation samples for model evaluation. UAV multispectral imagery was processed to obtain canopy reflectance information at the breeding plot scale, and the inversion performance of the method based on the 3D RTM was compared with that of a conventional data-driven empirical model.The results showed that the method based on the 3D RTM achieved stable and accurate LAI estimation under different experimental conditions. The overall inversion accuracy reached an R² of 0.80 with an RMSE of 0.98, outperforming the empirical model (R² = 0.78 and RMSE = 1.04). In addition to improved estimation accuracy, the method also exhibited better temporal consistency and stronger robustness across different sites, growth stages, and wheat varieties. Incorporating three-dimensional canopy structural information effectively reduced the instability caused by canopy heterogeneity and improved the generalization capability of the inversion model.Overall, the proposed approach enables accurate and high-throughput estimation of winter wheat LAI at the breeding plot scale without relying on large quantities of ground training samples. The study demonstrates the potential of integrating 3D RTM with UAV multispectral remote sensing for crop phenotyping applications and provides reliable technical support for precision breeding and large-scale agricultural monitoring.