Multivariate Outlier Detection in Postprocessing of Multi-temporal PS-InSAR Results using Deep Learning

Abstract Multi-temporal InSAR (MT-InSAR) techniques proved to be very effective for deformation monitoring. However, decorrelation and other noise sources, can be limiting factors in MT-InSAR. The obtained observations (PS - Persistent scatterers) are usually very demanding from a computational perspective, as they can reach hundreds of thousands of observations. To simplify and speed up the classification process, in this study we present an approach based on Convolutional Neural Networks (CNN) classification models, for the detection of MT-InSAR outlying observations. For each PS, the corresponding MT-InSAR parameters, its neighbouring scatterers parameters and its relative position are considered. Tests in two independent PS datasets, covering the regions of Bratislava city and the suburbs of Prievidza, Slovakia, were performed. The results showed that such models are robust and reduced computation time method for the evaluation of MT-InSAR outlying observations. However, the applicability of these models is limited by the deformation pattern in which such models were trained.