Leaf angle distribution (LAD) can be used to describe the canopy structure of vegetation completely, such as crops, trees and grass. It’s one of the important parameters to quantitative description of vegetation canopy structure. At the present, there are few studies used the spectral data to inverse LAD, and results of the most existing studies of mean leaf tilt angle and leaf angle distribution were the locational inversion. Therefore, this study set the study site in five counties of Baoding City, Hebei Province, using terrestrial laser scanning (TLS) to acquire the leaf angle distribution data of maize. Combining the Landsat8 remote sensing data, firstly, the principle component analysis was taken to extract the principle information of measured leaf angle distribution of maize. Secondly, the back propagation artificial neural network was taken to model the relationship of principal information and spectral data. Then, the model was used in the whole study area to accomplish the upscaling transform. Finally, the upscaled mean tilt angel (MTA) was calculated based on the predicted LAD by principal component inverse transformation, in order to quantitate the leaf angle data. The cross validation result showed that the accuracy (R2) between upscaled MTA and measured MTA was 0.7862, and the mean square root error (RMSE) was 3.04°. Consequently, it shows that this method can realize the aim of LAD upscaling.