Premium
A numerical approach to the analysis and classification of channel network patterns
Author(s) -
Ichoku Charles,
Chorowicz Jean
Publication year - 1994
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/93wr02279
Subject(s) - channel (broadcasting) , trellis (graph) , pattern recognition (psychology) , parallelism (grammar) , confluence , bifurcation , type (biology) , artificial intelligence , contextual image classification , computer science , mathematics , image (mathematics) , algorithm , geometry , nonlinear system , parallel computing , geology , computer network , paleontology , decoding methods , physics , quantum mechanics , programming language
A large number of samples of visually classified channel networks are taken from some published works. The samples are representative of the following pattern types: dendritic, parallel, rectangular, trellis, and pinnate. By means of computer techniques, essential parameters of elements of these pattern samples are determined. These include lengths, directions, and degrees of curvedness and meandering of channel segments, as well as confluence angles. The parameters are used to determine pattern attributes such as density, texture, parallelism, rectangularity, and bifurcation ratios. Thresholds are generated for the attributes and used in the construction of classification models for the five pattern types studied. The processes and models are coded in a computer program for use in the automated classification of numerically valued channel networks. After classification the image is displayed with each individual network bearing a color which shows the pattern type to which it has been assigned.