A specific processing chain has been designed to build leaf chlorophyll content (LCC) prediction maps in banana fields using multispectral UAV imagery. A supervised regression model was first calibrated out-of-field using multispectral images of collected leaves laid on the ground, and then applied to build a LCC ortho-mosaic for a complete field using the same multispectral sensor. Though the quality of the resulting LCC map was limited mainly due to light scattering inside the canopy, a significant correlation with within-field vigor heterogeneity has been found, making this approach promising for further banana production management.