@InProceedings{DiazQuJiBeHaFePl:2015:FrSoAu,
author = "Diaz, Pedro Marco Achanccaray and Quirita, Victor Andres Ayma and
Jimenez, Luis Ignacio and Bernabe, Sergio and Happ, Patrick Nigri
and Feitosa, Raul Queiroz and Plaza, Antonio",
title = "SPT 3.0: A free software for automatic segmentation parameters
tuning",
booktitle = "Anais...",
year = "2015",
editor = "Gherardi, Douglas Francisco Marcolino and Arag{\~a}o, Luiz
Eduardo Oliveira e Cruz de",
pages = "5578--5581",
organization = "Simp{\'o}sio Brasileiro de Sensoriamento Remoto, 17. (SBSR)",
publisher = "Instituto Nacional de Pesquisas Espaciais (INPE)",
address = "S{\~a}o Jos{\'e} dos Campos",
abstract = "This paper presents a free software tool, named Segmentation
Parameter Tuner 3 (SPT 3.0), designed for automatic tuning of
segmentation parameters based on a number of optimization
algorithms using different quality metrics as fitness functions.
For a segmentation algorithm to produce segments that correspond
in some way to meaningful image objects, its parameters must be
properly tuned. Conventionally, it involves a long time consuming
series of trials-and-errors. Some initiatives towards designing
methods for automatic segmentation parameter tuning rely on a
stochastic optimization method. Basically, it searches the
parameter space for the values that maximize the level of
agreement between a set of reference segments, which are
delineated manually by a human operator, and the segmentation
outcome. This level of agreement is quantified by a metric which
compares the segmentation outcome with the reference segments
given by the user. As our target is to maximize the level of
agreement represent by this metric, it becomes an optimization
problem where the metric would be the fitness function. In this
version, SPT 3.0 offers many features such as: six segmentation
algorithms, which are able to work with Optical, Hyperspectral
and/or Synthetic Aperture Data (SAR) images (including parallel
GPU-based implementations for two of them), four alternative
optimization methods (Differential Evolution, Nelder-Mead, among
others) and seven different fitness functions (Hoover Index, Shape
Index, among others) are available, which assess the segmentation
outcome.",
conference-location = "Jo{\~a}o Pessoa",
conference-year = "25-29 abr. 2015",
isbn = "978-85-17-0076-8",
label = "1127",
language = "en",
organisation = "Instituto Nacional de Pesquisas Espaciais (INPE)",
ibi = "8JMKD3MGP6W34M/3JM4ECR",
url = "http://urlib.net/ibi/8JMKD3MGP6W34M/3JM4ECR",
targetfile = "p1127.pdf",
type = "Processamento de imagens",
urlaccessdate = "09 maio 2024"
}