Multitemporal change detection of urban trees using localized region-based active contours in VHR images

This study introduces active contours for multitemporal monitoring of urban trees using fine resolution imagery. In our implementation active contours locally identify tree crown image regions using a localized data energy term and competition of energy forces between adjacent contours. The simultaneous evolution of n contours represents an equivalent number of tree crowns which are initialized based on a fast and efficient image fitting procedure. We incorporate prior information into the multitemporal analysis, as optimized contours are propagated through a sequence of images. We demonstrate the applicability of the method to monitor individual trees and tree groups in two residential areas in the Netherlands using multispectral aerial images acquired over a period of five years. The method allowed the identification of abrupt and gradual tree crown changes in both areas and performed superior to an alternative region growing segmentation approach. We conclude that the method extracts meaningful image objects that facilitate tree crown monitoring and is effective to update tree crown spatial-databases in urban areas.

► We propose multitemporal change detection of urban trees based on active contours.
► The method successfully identified tree crowns in a time series of 6 VHR images.
► Contours were initialized semi-automatically and their evolving energies localized.
► Results were accurate and informative of abrupt and gradual changes.
► The method outperformed a multiresolution segmentation approach for tree monitoring.