Evolving forest fire burn severity classification algorithms for multi-spectral imagery

Created by W.Langdon from gp-bibliography.bib Revision:1.3973

@InProceedings{Brumby:2001:SPIE,
  author =       "S. P. Brumby and J. J. Bloch and N. R. Harvey and 
                 J. Theiler and S. Perkins and A. C. Young and 
                 J. J. Szymanski",
  title =        "Evolving forest fire burn severity classification
                 algorithms for multi-spectral imagery",
  booktitle =    "In Algorithms for Multispectral, Hyperspectral, and
                 Ultraspectral Imagery VII, Proceedings of SPIE",
  year =         "2001",
  editor =       "Sylvia S. Shen and Michael R. Descour",
  volume =       "4381",
  pages =        "236--245",
  keywords =     "genetic algorithms, genetic programming, Multispectral
                 imagery, Supervised classification, Forest fire,
                 Wildfire, GENIE, Aladdin",
  URL =          "http://public.lanl.gov/perkins/webdocs/brumby.aerosense01.pdf",
  DOI =          "doi:10.1117/12.437013",
  size =         "10 pages",
  abstract =     "Between May 6 and May 18, 2000, the Cerro Grande/Los
                 Alamos wildfire burned approximately 43,000 acres
                 (17,500 ha) and 235 residences in the town of Los
                 Alamos, NM. Initial estimates of forest damage included
                 17,000 acres (6,900 ha) of 70-100percent tree
                 mortality. Restoration efforts following the fire were
                 complicated by the large scale of the fire, and by the
                 presence of extensive natural and man-made hazards.
                 These conditions forced a reliance on remote sensing
                 techniques for mapping and classifying the burn region.
                 During and after the fire, remote-sensing data was
                 acquired from a variety of aircraft-based and
                 satellite-based sensors, including Landsat 7. We now
                 report on the application of a machine learning
                 technique, implemented in a software package called
                 GENIE, to the classification of forest fire burn
                 severity using Landsat 7 ETM+ multispectral imagery.
                 The details of this automatic classification are
                 compared to the manually produced burn classification,
                 which was derived from field observations and manual
                 interpretation of high-resolution aerial
                 colour/infrared photography.",
  notes =        "p3 Max size directed acyclic graph, not tree GP. GENIE
                 object-oriented Perl. RSI's IDL language and image
                 processing environment. C. UNIX Linux.

                 Aladdin JAVA.

                 Output is written to one of a number of scratch planes
                 (memory) 'temporary workspaces where an image plane can
                 be stored.' 'the gene [ADDP rD0 rS1 wS2] applies
                 pixel-by-pixel addition to two input planes, read from
                 data plane 0 and from scratch plane 1, and writes its
                 output to scratch plane 2.' 'GENIE performs an analysis
                 of chromosome graphs when they are created and only
                 carries out those processing steps that actually affect
                 the final result. Therefore, the fixed length of the
                 chromosome acts as a maximum effective length.'

                 hamming distance fitness

                 pop=50. 30 gens. max chrome size 20. 3 scratch
                 registers.

                 'The best evolved image-processing algorithm had the
                 chromosome, [OPEN rD1 wS1 1 1][ADDS rD4 wS3 0.34][NEG
                 rS1 wS1][MULTP rD4 rS3 wS2] [LINCOMB rS1 rD6 wS3
                 0.11][ADDP rS1 rS3 wS1][SUBP rS1 rD5 wS1]'

                 'The final values of S1, S2, and S3 are then combined
                 in the linear sum, where the coefficients and intercept
                 have been chosen by the Fisher discriminant, as
                 described in Section 2.3, above, to produce our
                 real-valued answer plane A (Figure 6): A = 0.0147*S1 -
                 0.0142*S2 + 0.0134*S3 + 1.554'

                 'Adjusting the threshold on A to fall at the
                 between-peak minimum of the histogram at 0.7930 (a
                 different optimisation criterion for the threshold than
                 that used by default by GENIE) produces a new Boolean
                 mask, Figure 9, in which almost all the false positives
                 have been removed, and the remaining pixels marked as
                 burn correspond very closely to the high severity burn
                 regions in the BAER map'",
}

Genetic Programming entries for Steven P Brumby Jeffrey J Bloch Neal R Harvey James Theiler Simon Perkins A Cody Young John J Szymanski

Citations