1329
IS05725-2 STANDARD APPLICATION TO VERIFICATION OF ORTHOPHOTO-BASED
IMPERVIOUS SURFACE AREA AND IMPERVIOUSNESS FACTOR DETERMINATION
B. Hejmanowska, W. Drzewiecki, A. Wrobel
Dept, of Geoinformation, Photogrammetry and Remote Sensing of Environment, AGH University of Science and
Technology, Krakow, Poland - (galia, drzewiec, awrobel)@agh.edu.pl
KEY WORDS: Imperviousness Factor, Photointerpretation, Accuracy, IKONOS, Airborne Ortophotomap
ABSTRACT:
The main aim of the research described in the paper was analyse the accuracy of photointerpretation of impervious surface using
IKONOS images and its influence on the imperviousness factor determination. Two kinds of IKONOS image were chosen to tests:
panchromatic and colour pansharp. Airborne ortopho (pixel size of 0.2m) was applied as a reference. Six operators, digitised three
times, two kinds of IKONOS image, on the six test areas (300x300m). Accuracy analysis was performed applying different
parameters, among others: RMS and reproducibility (ISO 5725-2). Then, each test area was grided with 30m pixel size (simulation
Landsat image) and imperviousness factor was in each pixel determined. Mean error for PAN image was ca. 20% and for RGB
image ca. 10%.
1. INTRODUCTION
In many cases traditional land-use / land cover map created
through classification of satellite images does not provide us
with information necessary for evaluation of changes occurring
in the landscape. The process of landscape urbanization can be
given as an example. In this case the changes may be twofold.
What can change is not only the type of land-use (eg. from
agriculture to residential area), but also the level of urbanisation
within the same land-use type. For former the traditional land-
use map is enough, for latter may not. Because what changes
here it is not the land-use type, but the proportions of different
kinds of land cover inside the same land-use class. Very
detailed land-use / land cover map made from high resolution
satellite images or air photos could be proposed in such a case,
but such a map is very laborious and expensive when large area
is taken into consideration. Moreover, this kind of images may
be not available for past years. Continuum-based classification
of medium-resolution satellite images may be seen as a viable
alternative (see e.g. Clapham 2003, Xian and Crane 2005, Xian
2006). As a result of such classification a map of
imperviousness factor is obtained. The imperviousness factor
can be defined as a percentage of the area (e.g. percentage of
the image pixel) covered by impervious surfaces (such as roofs,
asphalt roads, parking lots, etc.).
Medium resolution satellite images have been used for the
assessment of the ground surface imperviousness from 1970s
(see Jackson 1975). Initially the methodology was based on
supervised or unsupervised image classification techniques, but
because of the resolution of these images the results were often
not satisfactory. Then many new approaches have been
developed, including among others artificial neural network,
spectral mixture analysis or regression tree approach. A review
of up-to-date techniques can be found e.g. in Weng (2008). The
accuracy of the imperviousness factor estimation reported in
different studies is usually better then 20 per cent.
Regardless the approach applied, the information about the
impervious surfaces acquired in the field or from higher
resolution data is needed as a training (or calibration) data and
also for accuracy assessment. The field data are rarely available
and in the most cases such information is acquired from digital
aerial or satellite orthophotos. High resolution satellite images
are commonly used for this purpose. Here we comes to the
question about the accuracy of these training and more
importantly validation (or control) datasets. In many cases high
resolution satellite orthoimages are used. Despite our efforts we
weren’t able to find in literature any assessments of accuracy
for imperviousness factor estimations based on photo
interpretation of high resolution satellite imagery. Actually the
only information about the accuracy of photo interpretation
based imperviousness data was find in Deguchi and Sugio
(1994). They use the aerial photographs in different scales
(from 1:10000 to 1:23000) to obtain the reference dataset. They
report the accuracy of the estimation of imperviousness factor
by visual interpretation of these photographs to be about 10 per
cent. We could expect similar or even worse accuracy of
imperviousness factor derived by visual interpretation of high
resolution satellite imagery. The verification of this assumption
was set as a goal in research presented in our paper.
2. ISO 5725-2 STANDARD AND ITS APPLICATION TO
PHOTO INTERPRETATION
Acquisition of spatial data should be accompanied by
acquisition of information about their quality. In our opinion
information about GIS data accuracy should be seen as one of
the most important metadata, especially if the data are to be
used in financial context as penalty (e.g. Integrated
Administration Control System - IACS, in agriculture financial
subsidies in EU) or taxation (e.g. cadastre, sewer waters). A
necessity for such information is also stressed in official
regulations. In the Directive of European Council from 14
March 2007 establishing the Infrastructure for Spatial
Information in the European Community (INSPIRE) we can
find the following statement: “metadata in spatial database shall
include information on the quality and validity of spatial data
sets” (Chapter II, Metadata, Article 5, p.2 c). GIS data metadata
as defined in ISO 19113 standard contain among others: quality,
spatial accuracy, temporal accuracy and thematic accuracy.