Image 100 Classification Methods for Erts Scanner Data

Abstract

The IMAGE 100 is a hardware-software system developed for analysis of digital imagery by the General Electric Corporation under contract to the Canada Centre for Remote Sensing (CCRS). The CCRS IMAGE 100 includes the following hardware: DEC PDP 11/40 computer with 72 thousand words of memory; 44 million word disk drive; two 800/1600 BPI 125 ips tape drives; 2 million word disk cartridge, 10 million bit solid state memory; high speed image analyzer (IMAGE 100); line printer/plotter. This system is controlled by a complex collection of FORTRAN and Assembler programs operating under a multi-tasking monitor The primary purpose of the CCRS IMAGE 100 is to permit investigators across Canada to be able to analyse easily and rapidly the multi-spectral scanner data of ERTS (Earth Resources Technology Satellite). This paper describes the mathematical methods and system procedures used to classify scanner data with the CCRS IMAGE 100. The system is capable of supervised or unsupervised classification for five channel input data with parametric or non-parametric distributions. Advantage is taken of the multilevel classification system to reduce the number of inter-class decisions, and thus speed the classification process. Results of classifications for the various methods are presented for a typical scene.