@Article{MausCâSoSaRaQu:2016:TiDyTi,
author = "Maus, Victor Wegner and C{\^a}mara, Gilberto and Souza, Ricardo
Cartaxo Modesto de and Sanchez, Alber Hermersson Ipia and Ramos,
Fernando Manuel and Queiroz, Gilberto Ribeiro de",
affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {Instituto
Nacional de Pesquisas Espaciais (INPE)} and {Instituto Nacional de
Pesquisas Espaciais (INPE)} and {Instituto Nacional de Pesquisas
Espaciais (INPE)} and {Instituto Nacional de Pesquisas Espaciais
(INPE)} and {Instituto Nacional de Pesquisas Espaciais (INPE)}",
title = "A time-weighted dynamic time warping method for land-use and
land-cover mapping",
journal = "IEEE Journal of Selected Topics in Applied Earth Observations and
Remote Sensing",
year = "2016",
volume = "9",
number = "8",
pages = "3729--3739",
month = "Aug.",
keywords = "Dynamic programming, image sequence analysis, monitoring, pattern
classification, time series.",
abstract = "This paper presents a time-weighted version of the dynamic time
warping (DTW) method for land-use and land-cover classification
using remote sensing image time series. Methods based on DTW have
achieved significant results in time-series data mining. The
original DTW method works well for shape matching, but is not
suited for remote sensing time-series classification. It
disregards the temporal range when finding the best alignment
between two time series. Since each land-cover class has a
specific phenological cycle, a good time-series land-cover
classifier needs to balance between shape matching and temporal
alignment. To that end, we adjusted the original DTW method to
include a temporal weight that accounts for seasonality of
land-cover types. The resulting algorithm improves on previous
methods for land-cover classification using DTW. In a case study
in a tropical forest area, our proposed logistic time-weighted
version achieves the best overall accuracy of 87.32%. The accuracy
of a version with maximum time delay constraints is 84.66%. A
time-warping method without time constraints has a 70.14%
accuracy. To get good results with the proposed algorithm, the
spatial and temporal resolutions of the data should capture the
properties of the landscape. The pattern samples should also
represent well the temporal variation of land cover.",
doi = "10.1109/JSTARS.2016.2517118",
url = "http://dx.doi.org/10.1109/JSTARS.2016.2517118",
issn = "1939-1404 and 2151-1535",
language = "en",
targetfile = "maus_a time.pdf",
urlaccessdate = "01 maio 2024"
}