A Fully Automated Approach to Classifying Urban Land Use and

Report
A Fully Automated Approach to
Classifying Urban Land Use and Cover
from LiDAR, Multi-spectral Imagery,
and Ancillary Data
Jason Parent
Qian Lei
University of Connecticut
Land cover and land use

Land cover: the physical material
on the earth’s surface (e.g. water,
grass, asphalt, etc.)

Land use: the use of the land by
humans (e.g. reservoir, agriculture,
parking lot, etc.)

Fundamental to landscape analyses
and urban planning.
2
Opportunities and challenges for
high resolution data
 Increasing availability of airborne light detection and
ranging (LiDAR) and aerial imagery offers
opportunities to study landscapes in great detail.
 Technically challenging to process…



require lots of hard drive space.
datasets must be divided into small subsets for processing.
conventional algorithms not well suited to processing large
numbers of subsets
Study objectives and justification

Develop fully automated algorithm to classify high
resolution (1-meter) land cover / land use which is
applicable over large areas.


Specifically, we developed python scripts with
ArcGIS to…


4
no previously presented algorithm has been feasible to
apply over large areas.
classify 1-meter land cover from LiDAR and
multispectral data.
infer land use from object geometry and spatial context
of land cover features.
Study area

Located in eastern Connecticut
in the northeastern U.S.

Semi-random stratified sample
of 30 1x1 km tiles.

5
Stratified by % impervious
cover (according to
Connecticut’s Changing
Landscape land cover data).
4800 km2
% impervious
0 - 33
33 - 66
66 – 100
Data
LiDAR




2010 leaf-off fall acquisition
Small footprint (44 cm)
Near-infrared (1064 nm)
> 1.5 pts/m2
6
Aerial orthophotos



2012 leaf-off spring
acquisition
Blue, green, red, and NIR
0.3 meter resolution
Land cover classification rules
Land cover
Building
Low impervious cover
(low IC)
Deciduous forest
Coniferous forest
Medium vegetation
Water
Riparian wetlands
Low vegetation
7
Primary characteristics
Height > 2.5m; no ground returns
Low NDVI; no returns 2 to 4.5
meters above ground
Height > 3m; high NDVI
Pixel- and object-based
Height > 3m; very high NDVI
rules
using
structural
and
Height 0.5 to 3m; high NDVI
spectral properties
No returns
Low reflectance in all bands; adjacent
to water
High return intensity
Land cover classification example
Land Cover
deciduous
trees
deciduous
evergreen
trees
coniferous
medium
med.vegetation
veg.
grass
vegetation
low/ low
veg.
water
water
wetland
wetland
building
building
road
low/ parking
IC lot / barr
8
Land cover class accuracies


User accuracy: probability that a cell label is correct.
Producer accuracy: probability that a cell is correctly labelled.
Class
Water
Building
Low vegetation
Wetland
Low impervious
Med. vegetation
Coniferous trees
Deciduous trees
User acc. (%) Prod. acc. (%)
96
99
91
26
93
61
90
95
85
97
94
35
91
60
76
96
n = 3200
93%
overall
Kappa =
0.90
Land use classification rules
Building use
Primary characteristic
Non-Residential
Large parking area; flat roof; large
Objectand parcelbuilding
size
based
rules
Large parking
area;using
narrowobject
building
width; similar building shapes
shape/size and parcel
Small parking area; peak roof;
landbuilding
cover
small
sizecomposition
Multi-family residential
Single family residential

10
Parcel cadastral information not used because of
limited availability.
Land use preliminary results
Land Cover
deciduous
trees
deciduous
evergreen
trees
coniferous
medium
med.vegetation
veg.
grass
vegetation
low/ low
veg.
water
water
wetland
wetland
building
building
Land
Use
road
low/ parking
IC lot / ba
18TBG5110_bldg
ParcelCls
Multi-Family
Residen
multi-family
Non-Residential
non-resid.
Single-Family
Reside
single-family
Land use classification assessment

Qualitative assessment
notes…


12
small commercial
buildings misclassified as
single family due to
similar structural
characteristics
problems caused by
mismatch between land
cover and parcel data
12
Conclusions and future work

Land cover classification:



Land use classification:


13
Use of airborne LiDAR and multi-spectral data proved
highly effective in classification of high resolution land
cover.
Developed fully automated algorithm that performs well
over large area.
Use of building shape and context is promising
Future work will develop rules for classification of…
 roads vs. parking lots
 urban vs. non-urban forest
 agriculture vs. turf
Questions?
A Fully Automated Approach to
Classifying Urban Land Use and
Cover from LiDAR, Multi-spectral
Imagery, and Ancillary Data
Jason Parent
Qian Lei
University of Connecticut

similar documents