ZUSAMMENFASSUNG
Erfolgreiche Experimente in der Landkartenherstel 1 unq städtischer Boden
nutzung, durch Verwendung algorithmischer Mustererkennung in der Digital -
Kalkuliermaschine, indem Aufnahmen von multispektralen Uberblicksdaten vom
"Earth Resources Technology Satel 1 ite-1," gemacht wurden, sind von Forschern
des geographischen Programms, U S Geological Survey, und dem Labor für
Anwendung von "Remote Sensing," Purdue University, durchgeführt worden. Als
Format der Karten erscheint ein Kalkuliermaschinen-Liniendruck, sowie Farb
drucke von Filmaufnahmen. Zahlreiche Kategorien von Bodennutzungsund Boden
bedeckungsformen wurden identifiziert und Kartographiert, um vollen Gebrauch
des Potenziells des Kalkuliermaschinen-Überblickssystems zu machen. Einige
Neuerungen in maschineller Analyse wurden entwickelt in Antwort auf: 1)
Unterscheidung von ländlichen und städtischen Gebieten; 2) Anordnung von
Daten in Kartenformat; 3) Zusammenstellung von Resultaten durch statistische
Einheiten; und 4) Lösung von weitern Problemen. Anwendung und Kritik der ent
wickelten Methodik zeigen Fortschritt, identifizieren Schwächen des Systems
und deuten auf Richtungen für weitere Verbesserungen.
INTRODUCTION
Since the launching of Earth Resources Technology Satellite-1 and follow
ing the shutdown of its vidicon system, considerable effort has been made to
utilize data received from the satellite's multispectral scanner system. Whi1
most of the total number of applications have used the visual images recon
structed from digital data by NASA in both individual bands and in false color
the computer compatible digital tape data have also been heavily used. These
data have been processed by digital computer, employing pattern recognition
algorithms, to examine spatial distributions by making "computer-drawn" maps
and other graphic displays. Topics as diverse as soils, vegetation, land-
forms, crops, and land use have been studied. Of the latter, both rural and
urban land use have been mapped and measured in numerous experiments.
The goal of the research reported on in this paper is to discover how
computer-aided processing of ERTS-acguired multispectral data can be employed
to map urban land use in the best possible way given current levels of com
puter technology and the relatively simple configuration of the satellite's
scanner. Attempts have been made to map these land uses with the highest
degrees of accuracy possible while delineating the maximum practical number
of categories. Further, the highest locational accuracy standards possible
have been attempted with the computer-generated graphics.
Numerous researchers have made advances in the field within the last
two plus years since the launch of ERTS-1. Centers equipped with computer
facilities, often previously used to process aircraft-borne scanner infor
mation, made early thrusts in this endeavor. Todd (1973) and others at the
Laboratory for Applications of Remote Sensing (LARS), Purdue University
found -- in an experimental mapping of the land use of Milwaukee, Wiscon
sin -- the urban scene to be more complex spectrally than it was concep
tually to the photo interpreter. Residential and commercial areas alike
