Séparation de sources thermométriques pour la détection de fuites dans les digues en terre

National audience; DTS (Distributed Temperature Sensors) based on the optical fibers present an economical efficient solution for the detection of water leakages in the dikes of canal. However, the acquired temperature data are corrupted by factors like daily and seasonal variations, geophysical conditions, noise, etc. The signal processing techniques must thus be applied to extract the useful information linked to the anomalies. In this context, the leakage detection can be modeled as a source separation problem. The source separation techniques of SVD and Independent Component Analysis are applied to the data. The self-response of the ground in which fiber is buried is the most coherent component of the data and masks the information linked to anomalies. This response is estimated as the first source of SVD. ICA is then applied on the residual sub-space obtained thus allowing to localize the leakages. The proposed method is validated on real data for the detection of artificial percolation type and real leakages.