Description
Date depot: 1 janvier 1900
Titre: AUTOMATIC DETECTION OF STRUCTURED ELEMENTS IN METER RESOLUTION EO OPTICAL IMAGES
Directeur de thèse:
Mihai DATCU ()
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé:
Earth Observation optical images may contain distortions or artificial structures which come from the instrument itself or which are due to their processing. These perturbations may come from an inadequate sampling schema, an incorrect filtering process, calibration residues, etc. They make image analysis more difficult and can decrease the efficiency of some algorithms because information is too distorted: for instance, some models may interpret these defects as significant textures, leading to classification errors. Thus, the PhD thesis will aim at developing parameter free methods able to automatically detect artefacts in EO images. The corresponding research will rely on two complementary approaches: an 'objective' one where complexity measures (intrinsic complexity and spatial variability) will be applied on images to characterize the defect signatures, and a 'subjective' one based on machine learning from expert knowledge.
Doctorant.e: Roman Gonzalez Avid