Premium
Selection of Less Biased Threshold Angles for SAM Classification Using the Real Value–Area Fractal Technique
Author(s) -
Shahriari Hadi,
Ranjbar Hojjatollah,
Honarmand Mehdi,
Carranza Emmanuel John M.
Publication year - 2014
Publication title -
resource geology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.597
H-Index - 43
eISSN - 1751-3928
pISSN - 1344-1698
DOI - 10.1111/rge.12042
Subject(s) - advanced spaceborne thermal emission and reflection radiometer , hydrothermal circulation , fractal , multispectral image , threshold limit value , selection (genetic algorithm) , geology , fractal dimension , mineralogy , remote sensing , mathematics , computer science , artificial intelligence , chemistry , seismology , mathematical analysis , organic chemistry , digital elevation model
The accuracy of classification of the S pectral A ngle M apping ( SAM ) is warranted by choosing the appropriate threshold angles, which are normally defined by the user. Trial‐and‐error and statistical methods are commonly applied to determine threshold angles. In this paper, we discuss a real value–area ( RV–A ) technique based on the established concentration–area ( C–A ) fractal model to determine less biased threshold angles for SAM classification of multispectral images. Short wave infrared ( SWIR ) bands of the A dvanced S paceborne T hermal E mission and R eflection R adiometer ( ASTER ) images were used over and around the S ar C heshmeh porphyry C u deposit and S eridune porphyry C u prospect. Reference spectra from the known hydrothermal alteration zones in each study area were chosen for producing respective rule images. Segmentation of each rule image resulted in a RV – A curve. Hydrothermal alteration mapping based on threshold values of each RV – A curve showed that the first break in each curve is practical for selection of optimum threshold angles. The hydrothermal alteration maps of the study areas were evaluated by field and laboratory studies including X –ray diffraction analysis, spectral analysis, and thin section study of rock samples. The accuracy of the SAM classification was evaluated by using an error matrix. Overall accuracies of 80.62% and 75.45% were acquired in the S ar C heshmeh and S eridune areas, respectively. We also used different threshold angles obtained by some statistical techniques to evaluate the efficiency of the proposed RV – A technique. Threshold angles provided by statistical techniques could not enhance the hydrothermal alteration zones around the known deposits, as good as threshold angles obtained by the RV – A technique. Since no arbitrary parameter is defined by the user in the application of the RV‐A technique, its application prevents introduction of human bias to the selection of optimum threshold angle for SAM classification.