COMBINING CANOPY MAPPING AND TREE DETECTION FOR URBAN TREE HERITAGE MONITORING FROM VHR IMAGES

Monitoring the tree heritage of large urban areas is a major new challenge for public decision-makers. The need for highly precise spatial information and annual updates, makes satellite and aerial remote sensing a preferred tool. In this article, we present the limits of two popular approaches: (1) canopy mapping by supervised classification of time series of satellite images at the pixel level and (2) automatic tree detection by DeepLearning approach. Then, we propose a strategy to combine the potential of these two approaches in an operational and reproducible system for monitoring urban tree heritage with very high precision.