Object-based change detection: tracking landscape dynamics with objects

Funding agency: Bijzonder Onderzoeks Fonds (BOF)

Research associate: Bos Debusscher

Promoters: Frieke Van Coillie

Acronym: OBCD

Duration: 5/2017 – 5/2021


Project objectives

Characterizations of land cover dynamics are among the most important applications of Earth Observation data, providing insights that inform management, policy and science.  In this project we abandon the per-pixel paradigm and apply change detection with object-based image analysis.

In spite of its promises, the development of object-based change detection monitoring techniques still faces numerous methodological challenges.  As yet, the most pressing questions are: (1) how to quantify changes between image objects, and (2) how to evaluate the accuracy of object-based change results?  This project will answer these questions by evolving innovative object-based metrics for change detection, and by developing new measures to assess the accuracy of objects and their change status.  Algorithmic developments will be initiated using change detection benchmark data (e.g. video sequences). Subsequently, the algorithms will be applied on multi-temporal real-world image sets covering both natural and anthropogenic landscapes, and acquired by different sensors, whereby ESA EO data from the Sentinels are particularly envisaged.

This project has three main objectives:

(1) establishment of an online OBCD benchmark database,

(2) development of object-based metrics for change identification, and

(3) development of new measures to assess the accuracy of objects and their change status.