@article{fdi:010093211,
  title = {{L}inear vs non-linear learning methods - {A} comparative study for forest above ground biomass, estimation from texture analysis of satellite images},
  author = {{T}apamo, {H}. and {M}fopou, {A}. and {N}gonmang, {B}. and {C}outeron, {P}ierre and {M}onga, {O}livier},
  editor = {},
  language = {{ENG}},
  abstract = {{T}he aboveground biomass estimation is an important question in the scope of {R}educing {E}mission from {D}eforestation and {F}orest {D}egradation ({REDD} framework of the {UNCCC}). {I}t is particularly challenging for tropical countries because of the scarcity of accurate ground forest inventory data and of the complexity of the forests. {S}atellite-borne remote sensing can help solve this problem considering the increasing availability of optical very high spatial resolution images that provide information on the forest structure via texture analysis of the canopy grain. {F}or example, the {FOTO} ({FO}urier {T}exture {O}rdination) proved relevant for forest biomass prediction in several tropical regions. {I}t uses {PCA} and linear regression and, in this paper, we suggest applying classification methods such as k-{NN} (k-nearest neighbors), {SVM} (support vector machines) and {R}andom {F}orests to texture descriptors extracted from images via {F}ourier spectra. {E}xperiments have been carried out on simulated images produced by the software {DART} ({D}iscrete {A}nisotropic {R}adiative {T}ransfer) in reference to information (3{D} stand mockups) from forests of {DRC} ({D}emocratic {R}epublic of {C}ongo), {CAR} ({C}entral {A}frican {R}epublic) and {C}ongo. {O}n this basis, we show that some classification techniques may yield a gain in prediction accuracy of 18 to 20%},
  keywords = {{CENTRAFRIQUE} ; {CONGO} ; {REPUBLIQUE} {DEMOCRATIQUE} {DU} {CONGO}},
  booktitle = {},
  journal = {{R}evue {A}fricaine de {R}echerche en {I}nformatique et {M}ath{\'e}matiques {A}ppliqu{\'e}es},
  volume = {18},
  numero = {},
  pages = {139--156},
  ISSN = {1638-5713},
  year = {2014},
  DOI = {10.46298/arima.1982},
  URL = {https://www.documentation.ird.fr/hor/fdi:010093211},
}