Temporal repetition detector for time series of spectrally limited satellite imagers

This article addresses the problem of estimating scene visibility in time series of satellite images. It focuses on satellites with few spectral bands and high revisit frequency. Our approach exploits the redundancy of information acquired during these revisits. It is based on an unsupervised algorithm that tracks local ground textures across time and detects ruptures caused mainly by opaque clouds and in some cases by haze, cirrus and shadows. Experiments have been carried out on 18 PlanetScope image times series of various locations. These time series come with hand-made ground truth labels that are published together with this paper. We compare our results with the Unusable Data Masks (UDM) that Planet provides together with the images, and demonstrate the effectiveness of the proposed method: success rates of 97.78% and 89.36% are reached for the visible and occluded regions classification. This article is related to the following publication: [Tristan Dagobert, Jean-Michel Morel, Carlo de Franchis and Rafael Grompone von Gioi, Visibility detection in time series of Planetscope images, IEEE International Geoscience And Remote Sensing Symposium, 2019].