Access to data with high spatial and spectral resolution is becoming more widespread and makes it possible to consider new applications for monitoring forest ecosystems. In particular, it is possible to consider studies of an entire stand but at the tree level. However, this raises questions about the joint use of data from different sensors such as LiDAR and hyperspectral imagers. This study presents a fusion methodology between high-density LiDAR data (45 pts/m² minimum) and VNIR hyperspectral images (HI) - (80 cm spatial resolution) acquired on french Alpine forests along an altitude gradient. The objective is to extract the main architectural characteristics of each individual tree and in particular the dimensions of crowns knowing the species. The methodology is based on the integration of HI and LiDAR data at different levels of fusion. First species are identified using the reflectance attributes contained in HI and the LiDAR canopy model, and then the 3D point cloud is segmented based on the allometric characteristics of the species. The integration of this additional information together with the segmentation algorithm provides an essential association strategy for LiDAR points located in the lower part of the canopy. Finally, the main dimensions of the crowns and associated trees are extracted from the 3D segmentation.