An Image-Pyramid-Based Raster-to-Vector
Conversion (IPBRTVC) Framework for
Consecutive-Scale Cartography and Synchronized
Generalization of Classic Objects
Chang Li, Xiaojuan Liu, and Wei Lu
Abstract
There are some key problems in
and cartographic generalization
automation and low accuracy in
vector conversion processing; (2) data-source inconsistency
in cartographic generation, i.e., different raster data sources
converted to vector; and (3) how to acquire arbitrary-scale
vector data. To solve these problems, we initially propose an
innovative image-pyramid-based raster-to-vector conversion
(
IPBRTVC
) framework with quality control for consecutive-scale
cartography and synchronized generalization, of which de-
tails can be modified accordingly under the
IPBRTVC
frame-
work. Landsat-8 imagery and Defense Meteorological Satellite
Program (
DMSP
)/Operational Linescan System (
OLS
) night-time
light imagery are used as a test dataset to extract classic
objects in the geometry level. Experimental results show that
the
IPBRTVC
framework not only solves the aforementioned
problems well but also (1) improves efficiency of data process-
ing by avoiding problems of corresponding features match-
ing and topology errors, (2) contributes to develop relevant
parallel computing system, and (3) helps to integrate the
raster-to-vector conversion and consecutive-scale cartography.
Introduction
The scale issue is known as one of the fundamental scientific
problems in remote sensing and cartography. The scale may
have variability, that is, targets at different scales will show dif-
ferent characteristics (Wu and Li 2009). The scale effect issue
has an important impact on the applicability of geographical
laws and algorithms deduced and induced by different scales.
The scale issue of remote sensing data lies in the diversity
of images with various spatial, spectral and temporal resolu-
tions. For example, the quality of spatial resolution reflects
the spatial details and the ability to extract the information of
the scene itself (Woodcock and Strahler 1987). Zhao and Bo
(2013) have enhanced the resolution of 500 m Moderate Reso-
lution Imaging Spectroradiometer (
MODIS
) reflectance images
and increased the spatial details of original
MODIS
images by
using super-resolution via sparse representation, which helps
ction for areas with a large number
ization presentation for multi-scale
plays a vital role in comprehensively
d features by mapping the 3D world
from a 2D image via remote sensing techniques. Using geo-
graphic object-based image analysis (
GEOBIA
) with very high
spatial resolution (
VHR
) aerial imagery (0.3 m spatial resolu-
tion) to classify vegetation, channel and bare mud classes in a
salt marsh, Kim
et al.
(2011) found that the multi-scale
GEOBIA
produced the highest classification accuracy, while the single-
scale approach produced an only moderately accurate classifi-
cation for all marsh classes. The image-pyramid-based method
(Adelson
et al.
1984), which is used to acquire multi-scale im-
ages with different resolutions by resampling and filter decom-
position, is the major method to realize the visualization pre-
sentation for remote sensing images. Combining the pyramid
transform and difference gradient detection algorithms is one
method to generate multi-scale linear decomposition images
(Simoncelli and Freeman 1995). It is practicable to improve
the pyramid feature detection accuracy by means of bringing
in steering coefficients to control the orientations of pyramid
transform according to the study (Greenspan
et al.
2002).
In cartography, the scale issue is seen as a main limiting
factor (Weibel 1997). Graphic elements in different spatial
scales are integrated with different mergeability and abstracta-
bility for properties. Li and Choi (2002) analyzed the associa-
tion of the change in the number of symbols or the percentage
of open space on similar scales with thematic attributes. They
studied road features on topographic maps of Hong Kong from
1:1000 to 1:200 000 considering six types of thematic attri-
butes, including “type”, “length”, “width”, “number of lanes”,
“number of traffic ways” and “connectivity”. More progress
has been made in the study of multi-scale representation in
thematic maps, such as drainage maps. The U.S. Geological
Survey (Brewer, Buttenfield, and Usery 2009), for instance,
has achieved multi-scale (from 1:200 000 to 1:100 000 000)
representation for national hydrological maps based on prior
knowledge, such as terrain, and climate. Four of these experi-
ments on basins show that corresponding prior knowledge has
instructive influence on cartographic generalization. In terms
of the scale transform, Buttenfield
et al.
(2010) has achieved
multi-scale (from 1:500 000 to 1:200 000) hydrographical
maps via cartographic generalization on the basis of the scale
of 1:240 000, which succeeds in the transformation from
Chang Li and Wei Lu are with the Key Laboratory for
Geographical Process Analysis & Simulation, Hubei Province,
and College of Urban and Environmental Science, Central
China Normal University, Wuhan, China, No. 152 Luoyu
Road, Wuhan, Hubei, 430079, P.R. China (lcshaka@126.
com (C.L.);
(X. L.); and
(W.L.)).
Xiaojuan Liu is with School of Geography and Planning, Sun
Yat-sen University, Guangzhou, China, No. 135 Xingang West
Road, Guangzhou, Guangdong, 510275, P.R. China
Photogrammetric Engineering & Remote Sensing
Vol. 85, No. 3, March 2019, pp. 169–178.
0099-1112/18/169–178
© 2019 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.85.3.169
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
March 2019
169
