A Weakly-Supervised Change Detection for Multispectral Earth Observation Imagery using a Long Short- Term Memory Classifier with a Virtual Training Data Neural Generator

This paper proposes a novel approach to improve accuracy of weakly-supervised change detection for multispectral Earth Observation (EO) imagery. The method is based on the idea to use an initial small-size EO labeled dataset to generate a larger set of virtual data. We have considered two variants of virtual data-generators based on the general architecture called Generative Adversarial Network (GAN): MLPGAN and LSGAN. The resulting virtual dataset is used to train a simple Long Short-Term Memory (LSTM) classifier. The proposed method is evaluated using the Mexico dataset acquired by the Thematic Mapper (TM) sensor of the Landsat 5 satellite. For each acquisition date, two spectral bands are considered (B4, B5). The two images have been acquired in April 2000 and May 2002, respectively. We have evaluated the change detection performances (OA, Kappa, MAR, and FAR) using two virtual data generators corresponding to considered GAN architectures. As a benchmark method, we have considered the case when the LSTM classifier is trained with the original small-size dataset without synthetic data generation.

I. Introduction

Change Detection (CD) is the technique that captures the spatial changes from multi-temporal remote sensing images acquired over the same geographical area due to manmade or natural phenomena [1]–[3]. Binary change detection represents the production of binary detection maps of changed and unchanged areas [1]–[7]. Due to advances in computing power and the increased number of satellites in orbit, there has been an increase in the interest taken in multi-temporal image analysis in recent years. [8] The possible applications of binary change detection done on images acquired over time are damage assessment after disasters, crop monitoring, detection of possible shifts in vegetation coverage, patterns of urban growth, and flood and fire control [9]–[11].