<oaidc:dc xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:oaidc="http://www.openarchives.org/OAI/2.0/oai_dc/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:dcterms="http://purl.org/dc/terms/" xmlns:marc="http://www.loc.gov/MARC21/slim">
<dc:source xsi:type="dcterms:URI">http://www.sudoc.fr/298673185</dc:source>
<dc:language xsi:type="dcterms:ISO639-2">fre</dc:language>
<dc:coverage xsi:type="unistra:Coverage">FR</dc:coverage>
<dc:title xsi:type="unistra:Titre" xml:lang="fre">Cartographie automatique de l&#8217;habitat informel &#224; Mayotte : &#233;valuation de l&#8217;apport des sources d&#8217;embeddings TESSERA &amp;amp; AlphaEarth Foundations</dc:title>
<dc:type xsi:type="unistra:Mention">M&#233;moire de master</dc:type>
<dc:description xsi:type="unistra:Discipline" xml:langue="fr">Observation de la Terre et G&#233;omatique</dc:description>
<dc:date xsi:type="unistra:Date">2026-06-30</dc:date>
<dc:description xsi:type="unistra:Resume" xml:langue="">Ce travail &#233;value l&#8217;apport des embeddings, des mod&#232;les de fondation TESSERA et AlphaEarth Foundations (AEF) pour cartographier automatiquement l&#8217;habitat informel &#224; Mayotte &#224; l&#8217;aide d&#8217;un algorithme de for&#234;ts al&#233;atoires (Random Forest). Les r&#233;sultats confirment la forte transf&#233;rabilit&#233; spatiale de ces mod&#232;les entra&#238;n&#233;s sur des sites distants, TESSERA se distinguant par sa grande sensibilit&#233; de d&#233;tection et par sa corr&#233;lation positive nette avec les zones expos&#233;es aux risques naturels majeurs. Malgr&#233; des verrous morphologiques, la raret&#233; des donn&#233;es de v&#233;rit&#233;-terrain et des lacunes g&#233;ographiques locales, cette &#233;tude d&#233;montre l&#8217;efficacit&#233; de ces technologies et pose les bases d&#8217;un suivi automatis&#233; annuel au service de la planification urbaine et de la justice sociale.</dc:description>
<dc:description xsi:type="unistra:Resume" xml:langue="">This work evaluates the contribution of embeddings from the TESSERA and AlphaEarth Foundations (AEF) foundation models to automatically map informal settlements in Mayotte using a Random Forest algorithm. The results confirm the strong spatial transferability of these models trained on distant sites, with TESSERA standing out for its high detection sensitivity and its clear positive correlation with areas exposed to major natural hazards. Despite morphological constraints, the scarcity of ground-truth data, and local geographical gaps, this study demonstrates the effectiveness of these technologies and lays the foundation for automated annual monitoring in the service of urban planning and social justice.</dc:description>
<dc:subject xsi:type="unistra:MotCle" xml:lang="fre">T&#233;l&#233;d&#233;tection</dc:subject>
<dc:subject xsi:type="unistra:MotCle" xml:lang="fre">Cartographie</dc:subject>
<dc:subject xsi:type="unistra:MotCle" xml:lang="fre">Urbanisme</dc:subject>
<dc:subject xsi:type="unistra:MotCle" xml:lang="fre">Bidonvilles</dc:subject>
<dc:subject xsi:type="unistra:MotCle" xml:lang="fre">Mayotte</dc:subject>
<dc:creator xsi:type="unistra:Auteur">Meyer, Sven</dc:creator>
<dc:contributor xsi:type="unistra:Directeur">Wenger, Romain</dc:contributor>
<dc:contributor xsi:type="unistra:AutreMembre">Salze, Paul</dc:contributor>
<dc:publisher xsi:type="unistra:Etablissement">Universit&#233; de Strasbourg</dc:publisher>
<dc:publisher xsi:type="unistra:CodeComposante">280732090</dc:publisher>
<dc:publisher xsi:type="unistra:Composante">Facult&#233; de g&#233;ographie et d'am&#233;nagement</dc:publisher>
<dc:publisher xsi:type="unistra:Laboratoire">Laboratoire image, ville et environnement</dc:publisher>
<dc:format xsi:type="dcterms:IMT">PDF</dc:format>
<dc:identifier xsi:type="dcterms:URI">https://publication-theses.unistra.fr/public/memoires/2026/GEO/Geographie_MEYER_Sven_2026.pdf</dc:identifier>
<dc:type xsi:type="unistra:Memoire">Memoire Unistra</dc:type>
<dc:rights xsi:type="unistra:Droits" xml:langue="fre">Acc&#232;s libre</dc:rights>
</oaidc:dc>