The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
1110 
consider combining the area-based matching (ABM) and the 
feature-based matching (FBM), matching parameter self-tuning, 
generation of more redundant matches and a coarse-to-fme 
hierarchical matching strategy. In particular, we also have to 
consider the fact that HRSI provides for some characteristics and 
possibilities for automatic image matching: 
(1) Compared to the traditional scanned 8-bit images, images 
from these sensors have better radiometric performance. Most of 
the linear array sensors have the ability to provide more than 
8-bit/pixel images. This results in a major improvement for 
image matching in terms of reducing the number of mismatches 
for “homogeneous” areas and especially for dark shadow areas. 
(2) Ability to provide multiple-view terrain coverage in one 
flight mission or satellite orbit. This enables the multi-image 
matching approach, which leads to a reduction of problems 
caused by occlusions, multiple solutions, surface discontinuities 
and results in higher measurement accuracy through the 
intersection of more than two image rays. Also, along-track 
stereo images, from the same orbit within a very short time 
interval, have a distinct advantage to those across-track because 
they reduce the radiometric differences, and thus increase the 
correlation success rate. 
In this paper, we present an advanced image matching approach 
and we will report about the key algorithms in details (Chapter 
2). We give a DTM accuracy evaluation using SPOT-5 
HRS/HRG triplets in a testfield in Zone of headstream of Three 
rivers, eastern Tibet Plateau, China. In another test, the proposed 
approach has been also applied to 23 IRS-P5 stereo pairs over 
Beijing city, the resulted 12.5 m DTM reproduced quite well not 
only the general features of the terrain relief but also small 
géomorphologie and other features visible in the IRS-P5 images. 
Through these experiments, we demonstrate that our approach 
leads to good results (Chpater 3). 
2. The Automatic DSM/DTM Generation Approach 
We have developed an advanced matching approach for 
automatic DSM/DTM generation from HRSI. It can provide 
dense, precise and reliable results. The approach uses a 
coarse-to-fme hierarchical solution with a combination of 
several image matching algorithms and automatic quality control. 
The new characteristics provided by HRSI imaging systems, i.e. 
the multiple-view terrain coverage and the high quality image 
data, are also efficiently utilized in this approach. 
The approach essentially consists of 3 mutually connected 
components: the image pre-processing, the multiple primitive 
multi-image (MPM) matching and the géomorphologie 
refinement matching procedure. The overall data flow is shown 
schematically in Fig. 1. The images and the given or previously 
estimated orientation elements are used as input. After 
pre-processing of the original images and production of the 
image pyramids, the matches of three feature types (feature 
points, grid points and edges) in the original resolution images 
are found progressively starting from the low-density features in 
the lowest resolution level of the image pyramid. A TIN form 
DSM is reconstructed from the matched features at each 
pyramid level by using the constrained Delauney triangulation 
method. This TIN in turn is used in the subsequent pyramid level 
for derivation of approximations and adaptive computation of 
some matching parameters. Finally and optionally, least squares 
matching methods are used to achieve more precise results for 
all matched features and for the identification of some false 
matches. 
In order to capture and model the detailed terrain features, our 
DSM/DTM generation approach not only generates a large 
number of mass points but also produces line features. Here we 
just give a detailed description about the core part of our 
approach, i.e. the Multiple Primitive Multi-Image Matching 
(MPM) matching procedure, for more details of this matching 
approach please refer to Zhang and Gruen, 2004,2006; Zhang, 
2005; Baltsavias, et. al., 2006. 
Fig. 1: Workflow of the proposed automated DTM/DSM 
generation approach. 
2.1 The Multiple Primitive Multi-image (MPM) Matching 
Procedure 
The Multiple Primitive Multi-Image (MPM) matching procedure 
is the core of our developed approach for accurate and robust 
DSM/DTM reconstruction. Results from this approach can be 
used as approximations for the refined matching procedure with 
least squares matching methods. In the MPM approach, the 
matching is performed with the aid of multiple images (two or 
more), incorporating multiple matching primitives - feature 
points, grid points and edges, integrating local and global image 
information and utilizing a coarse-to-fine hierarchical matching 
strategy. The MPM approach consists mainly of 3 integrated 
subsystems: the point extraction and matching procedure, the 
edge extraction and matching procedure and the relaxation based 
relational matching procedure. 
In the MPM matching procedure, we do not aim at pure 
image-to-image matching. Instead we directly seek for 
image-to-object correspondences. We have developed a new 
flexible and robust matching algorithm - Geometrically 
Constrained Cross-Correlation (GC 3 ) method in order to take 
advantage of the multiple images. The algorithm is an extension 
of the standard Cross-Correlation technique and is based on the 
concept of multi-image matching guided from object space and 
allows reconstruction of 3D objects by matching all available 
images simultaneously, without having to match all individual 
stereo-pairs and merge the results. 
2.2 Geometrically Constrained Cross-Correlation (GC 3 ) 
Algorithm 
We developed a new flexible and robust matching algorithm 
-GC 3 method in order to take advantage of the multiple images. 
The algorithm is an extension of the standard Cross-Correlation 
technique and is based on the concept of multi-image matching 
guided from object space and allows reconstruction of 3D 
objects by matching all available images simultaneously, without 
having to match all individual stereo-pairs and merge the results. 
Consider an IKONOS image triplet, as shown in Fig. 2. The 
middle image is chosen as the reference image and denoted as I 0 , 
the other two images are search images and denoted as I b i-1,2.