space is 
) bottom. 
a perfect 
» which 
umerical 
ive b* is 
nting the 
31 (SEII 
:d by the 
ally, only 
GB point 
ws (SEII, 
R 
;] © 
5 
ilue from 
a* varies 
120), b* 
e+ 120). 
(3) 
h can be 
and Z/Z, 
(4) 
luminant 
nt. Here 
).358037 
“ord and 
©) 
t Aa 
D* means 
pare the 
and CIE 
proach is 
on is that 
m visible 
ative for 
lose to - 
with CIE 
hreshold. 
  
The best threshold can again be obtained under human 
supervision. Meanwhile, it is of course not possible for NDVI 
to deal with RGB imagery. 
In order to get better removal result, a supervised learning 
method, support vector machine (SVM), is utilized for 
extracting vegetation. We choose (L, a, b) as characteristic 
variables. Given a set of training examples, each marked as 
belonging to one of two categories (vegetation or not), an SVM 
training algorithm builds a model that assigns new examples 
into one category or the other An SVM model is a 
representation of the examples as points in space, mapped so 
that the examples of the separate categories are divided by a 
clear gap that is as wide as possible (BURGES, 1998). Linear 
SVM can be used in this paper. 
An n-dimensional pattern (object) x has » coordinates, 
x=(x1, X3, ..., Xn), Where each x; is a real number, x; ER for i — 1, 
2, ...,n. Each pattern x; belongs to a classy; € {-1, +1}. 
Consider a training set 7 of m patterns together with their 
classes, T={(x;, V1), (X2,¥2), ---» (Xm Vm)}. Consider a dot 
product space S, in which the patterns x are 
embedded, x;, x», ..., x,, € $. Any hyperplane in the space S can 
be written as (BURGES, 1998) : 
{xeS|w-x+b=0}, weS,beR (6) 
The dot product wex is defined by: 
a) 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B3, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
    
wx-Y ux © 
i=l 
A training set of patterns is linearly separable if there exists at 
least one linear classifier defined by the pair (w, b) which 
correctly classifies all training patterns. This linear classifier is 
represented by the hyperplane H (wex+b=0) and defines a 
region for class +1 patterns (wex+b>0) and another region for 
class -1 patterns (wex+b<0). 
After training, the classifier is ready to predict the class 
membership for new patterns, different from those used in 
training. The class of a pattern x,is determined with the 
equation: 
+1 if w-x, +b>0 
class(x, ) = (8) 
if wx +h<0 
Therefore, the classification of new patterns depends only on 
the sign of the expression wex+b. After selecting samples, we 
can recognize two classes problem (vegetation or not) properly. 
3. EXPERIMENTS AND RESULTS 
The aim of this section is to evaluate the feasibility and 
effectiveness of the proposed occlusion detection technique. 
The proposed method was implemented by C++. 
3.1 CIR image segmentation and removal 
c) 
Figure 1. a) Standard false color composite satellite image. b) Result by NDVI. c) Result by CIE L*a*b. 
  
a) 
Figure 2. a) True color close-range image. 5) Result by CIE L*a*b with a threshold. c) Result by feature (L, a. b) with SVM. 
The first experimental data is satellite image that is composed 
by standard false color. And result of vegetation extraction by 
NDVI (with a threshold NDVI > 0.1) and CIE L*a*b (with a 
threshold a > 12) are assigned by green color in Fig 1. 5) and c). 
We can find that there are almost the same results between these 
two methods. So the CIE L*a*b approach is used for vegetation 
c) 
extraction in this paper. Furthermore CIE L*a*b, vegetation can 
also be extracted from visible light RGB images because the 
component a* is negative, which is tested as follows.