stem
any
racy
STATISTICAL TEST FOR EVALUATION OF THE ACCURACY OF DIGITAL MAPS FOR GEO-SPATIAL
INFORMATION SYSTEM
Prof. Bock-Mo Yeu
Dept. of Civil Engineering, Yonsei University
Prof. Hyun Kwon
Dept. of Civil Engineering, Suwon Technical University
Researcher Seok-Kun Lee
Dept. of Civil Engineering, Yonsei University
Researcher Dong-Bin Shin
Korea Research Institute for Human Settlements
Commission II, Working Group 5
KEY WORDS: Statistics, Production, Accuracy , Quality, Project, Geo Spatial Databases
ABSTRACT
Digital maps of Geo-Spatial Information System are very important as they are the foundation of the system and the necessity to
implement and develop digital map database is increasing . But because of various different needs and requirement of different
users, each has different needs of accuracy. A standard for statistical test for accuracy evaluation such as the hypothesis test is
very essential under such changing environment. In this study a statistical method for evaluating accuracy of digitized maps is
introduced . Using hypergeometric distribution and considering consumer risk and producer risk, a statistical method was used to
decide the optimal sample size.
1. INTRODUCTION
In Korea, with the development and growing interest of Geo-
Spatial Information System, the need for standards in digital
maps are being discussed and frameworks are being
established by government bodies such as the National
Geographic Institute. Meanwhile maps are being digitized,
trying to satisfy the standard as much as possible. It is well
known that the database implementation cost comprises most
of any GSIS projects and reports have been made that some
amounts up to 8096 of the whole project implementation
cost( K.Thaph, J.Bossler, 1992). Big cost for initial
production also implies big cost for editing and correction of
erroneous data and sometimes irreparable damages. All these
means that everything has to start with the correct data from
the initial stage. Despite this emphasis on the importance of
411
accurate data, the importance of the quality of data from the
producer and the consumer point of view is not getting enough
attention.
Topographic base maps are directly produced in digital format
nowadays by using analytical stereo plotters and test points are
selected from the produced maps and compared to ground
surveying results and residuals are computed and used as an
evaluation factor for the map. Although effective, due to extra
costs involved in the testing procedure producers tend to look
for other methods of evaluation. Also researches are done
recently to evaluate the classification accuracy of thematic
maps by statistical evaluation and sampling method using the
binomial distribution, considering the risk of both the
consumers and producers (Michael E, G., 1979) (G.H.
Rosenfield et. al., 1982) (S.Aronoff, 1982). Despite these
efforts more researches should be done on the evaluation of
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996