APPLICATIONS BASED ON ORTHOCORRECTED HIGH RESOLUTION AND 
HYPERSPECTRAL IMAGES 
P. Boccardo ^* ,E. Borgogno Mondino', F. Giulio Tonolo*? 
* Politecnico di Torino, Dipartimento di Georisorse e Territorio, C.so Duca degli Abruzzi 24, 10129 Torino, ITALY- 
(piero.boccardo, enrico.borgogno, fabio.giuliotonolo)@polito.it 
KEY WORDS: Remote sensing, Orthorectification, Archaeology, Mapping, Application, Updating 
ABSTRACT: 
The availability and variety of high resolution satellite and hyperspectral aerial images have led us to consider how image geometric 
aspects can condition remote sensing applications . Our main purpose is to underline the importance of images geometric correction; 
its quality conditions not only positioning, but also object dimensions measurements and shapes which can be very important in 
specific application as map updating and archaeological investigations, where some shapes could be buried. 
In this paper two orthoprojection procedure, which have been validated, are presented. Both of them are based on non-parametric 
self-developed algorithms (Rational Function Model and Neural Network) applied to different types of images aimed to determined 
their limits and potentialities correlated to their geometric features. Firstly map scale suitability of such data (which depends both on 
the geometric resolution of images and on the adopted sensor model) has been investigated through planimetric positioning accuracy 
tests. Presented experiences refer to orthoprojection of a SPOTS supermode image and to an airborne sensor MIVIS ( Multispectral 
Infrared Visible Imaging Spectrometer) one. 
Considerations have been then carried out about both geometric and content features of the obtained orthoimages. In particular 
SPOTS image has been used for demonstrating how well it can be used for middle scale map updating and how its scale mapping 
suitability heavily depends on the adopted geometric calibration method. MIVIS image has been used to underline how metric issues 
are as important as the spectral ones in the particular field of the archaeological investigation. In the first case attention has been, 
therefore, mainly paid to the geometric content of the SPOTS, while in the second one interpretation problem is also taken into 
consideration. 
1. ADOPTED GEOMETRIC CORRECTION METHODS 
Many applications are strictly dependent on the quality of 
geometry of images. Object shapes, dimensions and relation 
between objects are more than a simple positioning problem. 
Geometry correction has, therefore, a semantic importance that 
has not to be neglected. 
Geometric calibration of high resolution satellite/aerial images 
can be performed according to 2 different approaches: the 
rigorous and the non-parametric one. 
Rigorous models are based on time-dependent collinearity 
equations. Due to the common lack of detailed information 
about sensors and platforms these methods have been, for a 
long time, neglected by the users community. In their place 
generic methods (non-parametric) have indeed been developed. 
These are generic and independent both from the sensor and 
from the acquisition mode. 
Nowadays commercial software are equipped with both types, 
but often, user has no possibility of controlling the whole 
process. That's why we have decided to face directly the 
problem, developing by ourselves, the necessary routines. In 
this way we can really regards any deficiency of the methods. 
Attention has been particularly paid to the non-parametric 
methods. It's not purpose of this work to deeply understand 
these methods, but simply to give some basic information 
necessary to understand our working philosophy and to quantify 
residuals resulting from the orthocorrection operations. 
Results here presented have been obtained using two procedure 
developed in IDL and MATLAB programming languages. One 
performs the Rational Function Mode (RFM), perhaps the most 
famous generic 3D generic method. The other, particularly 
innovative, performs a neural network approach to the problem. 
All generic models require a high number of Ground Control 
Points (GCP). 
1.1 Neural Network Model 
Neural Network (NN) approach for geometric calibration 
purposes of remote sensing images can be considered an 
innovative and experimental solution. NN are mathematical 
models which simulate brain dynamics. Computational scheme 
can be thought as a flow of distributed information which are 
elaborated within computational node called "neurons" of the 
NN. Some of them (input) receive data from the external world 
, some give back information to it (output), some other simply 
communicate each other (hidden). Neurons are mathematically 
represented by weights, parameters of the model, which have to 
be estimated on the basis of the GCP through an iterative 
learning process. 
As far as this work is concerned the developed orthocorrection 
procedure is based on an opportunely designed Multi Layer 
Perceptron (MLP) NN. This type of NN has been chosen for its 
function approximation and estimation features. It shows its 
high suitability especially for non linear functions as considered 
relations are. Basic idea is to substitute the upward projecting 
model relating image (2,77) and ground (X,Y,Z) coordinate 
with a well designed and trained MLP NN. 
Reasons that have pushed us to consider this approach come 
from the will to solve recurrent problems associated to the RFM 
method described in the next paragraph. NN allow to avoid 
local linearisation of equations relating coordinates. They 
1067 
a.