@article{fdi:010091533,
  title = {{L}and cover classification of the {A}lps from {I}n{SAR} temporal coherence matrices},
  author = {{G}iffard-{R}oisin, {S}ophie and {B}oudaour, {S}. and {D}oin, {M}.{P}. and {Y}an, {Y}. and {A}tto, {A}.},
  editor = {},
  language = {{ENG}},
  abstract = {{L}and cover mapping is of great interest in the {A}lps region for monitoring the surface occupation changes (e.g. forestation, urbanization, etc). {I}n this pilot study, we investigate how time series of radar satellite imaging ({C}-band single-polarized {SENTINEL}-1 {S}ynthetic {A}perture {R}adar, {SAR}), also acquired through clouds, could be an alternative to optical imaging for land cover segmentation. {C}oncretely, we compute for every location (using {SAR} pixels over 45 × 45 m) the temporal coherence matrix of the {I}nterferometric {SAR} ({I}n{SAR}) phase over 1 year. {T}his normalized matrix of size 60, ×, 60 (60 acquisition dates over 1 year) summarizes the reflectivity changes of the land. {T}wo machine learning models, a {S}upport {V}ector {M}achine ({SVM}) and a {C}onvolutional {N}eural {N}etwork ({CNN}) have been developed to estimate land cover classification performances of 6 main land cover classes (such as forests, urban areas, water bodies, or pastures). {T}he training database was created by projecting to the radar geometry the reference labeled {CORINE} {L}and {C}over ({CLC}) map on the mountainous area of {G}renoble, {F}rance. {U}pon evaluation, both models demonstrated good performances with an overall accuracy of 78% ({SVM}) and of 81% ({CNN}) over {C}hamb{\'e}ry area ({F}rance). {W}e show how, even with a spatially coarse training database, our model is able to generalize well, as a large part of the misclassifications are due to a low precision of the ground truth map. {A}lthough some less computationally expensive approaches (using optical data) could be available, this land cover mapping based on very different information, i.e., patterns of land changes over a year, could be complementary and thus beneficial; especially in mountainous regions where optical imaging is not always available due to clouds. {M}oreover, we demonstrated that the {I}n{SAR} temporal coherence matrix is very informative, which could lead in the future to other applications such as automatic detection of abrupt changes as snow fall or landslides.},
  keywords = {},
  booktitle = {},
  journal = {{F}rontiers in {R}emote {S}ensing},
  volume = {3},
  numero = {},
  pages = {932491 [13 ]},
  ISSN = {2673-6187},
  year = {2022},
  DOI = {10.3389/frsen.2022.932491},
  URL = {https://www.documentation.ird.fr/hor/fdi:010091533},
}