X-B8, 2012 
bility that a differ- 
lues as large as or 
1g that the null hy- 
zer the probability 
1e measure values 
the same distribu- 
ones. 
ear in Table 1: in 
el was used for the 
dow of 5 x 5 pixels 
:ntropy was calcu- 
juantized in 8 bins 
LH2 quantization 
| the beginning. A 
| selection of 9 x 9 
in ‘LHR’ versions 
.LHRI and LHR2 
lation of the local 
13 pixel windows 
2 their dimensions 
ame parameters as 
vith the difference 
om the calculation 
ation invariant and 
ulated for radius 1 
d LBP4, the same 
Ito 2. The 'LTP' 
ective ‘LBP’ ones, 
ry system and the 
ctively. 
> t-tests are signif- 
ce usually used in 
). This provides a 
at the mean values 
hes are almost the 
ssumption that the 
antly smaller than 
1 specific measure 
istance can lead to 
erophyte patches. 
ned, it is observed 
y well and outper- 
riminatory power. 
band, seem to pro- 
ding the measures 
the modified local 
itperform all other 
appearing slightly 
he other hand, ap- 
ictory results even 
ations where the 
inner one, LHR2 
-test is concerned, 
ariant local ternary 
lified rotation vari- 
), LTBP4, both cal- 
order of magnitude 
:cordance with the 
. LTP1 and LTBP4 
it rates in correctly 
:d on their height. 
Jassification accu- 
y pattern approach 
   
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B8, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
Table 1: Evaluation of texture analysis measures through £-tests and classification. 
  
  
  
  
Blue band Green band Red band NIR band 
Method | CART | p-value | CART | p-value | CART | p-value | CART | p-value 
LE1 0.8429 | 1.66E-05 | 0.8571 | 2.03E-05 0.7 3.15E-08 | 0.6857 0.01 
LE2 0.7 7.43E-06 | 0.7429 | 5.09E-06 | 0.8286 | 8.09E-09 | 0.6857 0.0156 
  
LH1 0.7571 | 2.09E-19 | 0.7714 | 1.87E-16 | 0.6857 | 1.07E-04 | 0.6857 0.2447 
  
LH2 0.7286 | 4.67E-09 | 0.7714 | 3.40E-11 | 0.6429 | 1.75E-13 | 0.6857 | 7.09E-04 
  
LHR1 0.7286 | 2.50E-05 | 0.6857 | 1.32E-05 | 0.7143 | 3.50E-08 | 0.6857 | 2.31E-04 
  
LHR2 0.6857 | 1.94E-06 0.6 4.75E-07 | 0.7143 | 2.77E-09 | 0.6429 | 1.46E-04 
  
LHR3 0.6286 | 2.60E-03 | 0.6857 | 3.09E-05 | 0.6857 | 2.81E-07 | 0.6857 | 3.08E-03 
  
LHR4 0.6857 | 3.35E-06 | 0.6571 | 1.65E-07 0.7 2.53E-09 | 0.6857 | 1.89E-04 
  
LBP1 0.7571 | 8.71E-12 | 0.7429 | 4.02E-14 | 0.7714 | 2.67E-11 | 0.6857 0.8751 
  
LBP2 0.7714 | 1.53E-08 | 0.7571 | 7.57E-11 0.8 9.95E-08 | 0.6857 0.0652 
  
LBP3 0.7286 | 3.80E-08 | 0.7429 | 4.37E-09 | 0.6714 | 1.30E-07 | 0.6857 0.9499 
  
LBP4 0.6714 | 1.37E-03 | 0.7857 | 6.00E-04 | 0.6857 | 7.00E-03 | 0.6857 0.3399 
  
LTP1 0.8286 | 7.95E-17 | 0.9143 | 6.21E-19 | 0.7714 | 6.57E-16 | 0.6857 | 1.42E-02 
  
LTP2 0.8 2.66E-14 | 0.9143 | 3.80E-15 | 0.8143 | 2.04E-14 | 0.6857 | 2.69E-02 
  
LTP3 0.8143 | 2.42E-13 | 0.8143 | 2.91E-13 | 0.6571 | 9.75E-09 | 0.6857 0.0177 
  
LTP4 0.6857 | 2.88E-02 | 0.6857 | 5.20E-02 | 0.6857 | 5.71E-02 | 0.6857 0.6509 
  
LTBP1 | 0.7857 | 5.93E-14 | 0.8286 | 3.25E-15 | 0.8857 | 7.74E-15 | 0.6857 | 3.36E-03 
  
LTBP2 0.9 4.94E-16 | 0.9857 | 6.23E-17 | 0.8286 | 1.99E-16 | 0.6857 | 6.28E-03 
  
LTBP3 | 0.8571 | 7.09E-14 | 0.8714 | 1.14E-16 0.8 5.26E-18 | 0.6857 | 7.16E-03 
  
  
  
  
  
  
LTBP4 | 0.8286 | 2.32E-17 | 0.9429 | 1.52E-21 | 0.8286 | 1.56E-16 | 0.6857 0.0109 
  
  
  
  
  
  
with radius 1, LTBP2, having also a very low p-value and reach- 
ing an accuracy of 98.5796. The local variance instance of 3 x 3 
size window, LE1, shows a high classification rate in the green 
and blue bands and outperforms the local entropy, local entropy 
ratio and local binary patterns approaches. As previously, data 
from the green band seem to provide the best classification for 
almost all measures, apart from the local entropy ratio instances, 
for which better classification rates are achieved in the red band. 
Similarly to the t-tests, instances based on local ternary and the 
modified local binary patterns in the green band are the top per- 
forming ones, while data from the NIR band result in the lowest 
classification performance for almost all measures. 
In general, the lower the p-value of a method, the highest its clas- 
sification rate. Therefore, comparing two measures, the one with 
the lowest p-value is expected to provide the highest classifica- 
tion accuracy. However, as seen in Table 1, this general idea is 
not always true. This is caused by the random split of the data 
into training and test data, where the existence of outliers may 
influence the results of the t-test and classification to a different 
degree. 
5 CONCLUSIONS AND FUTURE WORK 
Considering a multispectral Quickbird image as the only source 
of data to discriminate between low and high vegetation habitats, 
à series of texture analysis measures, quantifying the degree of 
homogeneity of the texture, were proposed and evaluated. The 
approach is based on the idea that the shorter and smaller the 
vegetation, the more homogeneous the texture of the area will 
appear. On the contrary, in areas with tall vegetation, inhomoge- 
neous texture appears because vegetation canopy, tree trunks and 
ground alternate. 
Local variance, local entropy and local binary patterns served as 
the basis for the extraction of the proposed measures. Several 
Variations and different parameters were tested for each measure 
for all the available Quickbird bands. It was found that, in gen- 
eral, measures calculated from data from the green band outper- 
formed the ones from the other bands. As far as the methods 
are concerned, a modification of the local binary patterns ap- 
proach, assigning a 0 value to the pixels differing within a pre- 
defined range from the central pixel of the window under con- 
sideration, the value of 1 otherwise, seemed to outperform the 
other approaches in most bands. Among the tested instances of 
the method, the rotation variant one with a window radius equal 
to 1 calculated in the green band, was the most able to capture 
local texture variations and showed the best classification results. 
Some instances of another variation of the local binary patterns, 
forming ternary numbers, performed similarly well. 
In general, instances of different measures performed well in dis- 
criminating habitats with vegetation lower or higher than 2m. The 
results are promising for future extension to the discrimination 
of more height categories. The efficiency of the measures with 
other passive sensors can be part of future research tasks, as well 
as their application to other spectral bands or combinations of 
bands. The results encourage future research in texture analysis 
methods as alternatives in vegetation height estimation without 
the use of active sensors, such as LiDAR, or the need of exten- 
sive field campaigns. This can reduce the cost of land cover and 
habitat mapping through the use of less data and facilitate eco- 
logical monitoring and environmental sustainability planning. 
ACKNOWLEDGEMENTS 
The work presented was supported by the European Union Sev- 
enth Framework Programme FP7/2007-2013, SPA. 2010.1.1-04: 
“Stimulating the development of 490 GMES services in specific 
area”, under grant agreement 263435, project BIO_SOS: BIOdi- 
versity Multi-Source Monitoring System: from Space To Species, 
coordinated by CNR-ISSIA, Bari-Italy. 
REFERENCES 
Breiman, L., Friedman, J., Stone, C. J. and Olshen, R. A., 1984. 
Classification and Regression Trees. Chapman & Hall/CRC, 
Boca Raton, FL.