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
SIFT FOR DENSE POINT CLOUD MATCHING AND AERO TRIANGULATION
Jaan-Rong Tsay ® *, Ming-Shiuan Lee"
* Dept. of Geomatics, National Cheng Kung University, 70101 Tainan, Taiwan - tsayjr@mail.ncku.edu.tw
? Dept. of Geomatics, National Cheng Kung University, 70101 Tainan, Taiwan - evenif0901@msn.com
Commission III, WG III/1 /
KEY WORDS: SIFT, Dense Matching, Quality Filtering (QF), Aerotriangulation, Point Cloud
ABSTRACT:
This paper presents a new method for dense point cloud matching and aero triangulation based on the well-known scale invariant
feature transform (SIFT) technique. The modern digital cameras can take high resolution aerial images with high end lap between
contiguous images in a strip and, if needed, also with high side lap between images on neighboring strips. Therefore, automation on
image matching for generation of high density of 3D object points becomes applicable. A new method is thus developed to perform
the processing. Moreover, it can do an aero triangulation and automatic tie point measurement without the need on the input data
such as block and strip data for providing image overlap information. In order to increase the effectiveness of the method for
simultaneously processing a large number of aerial images with large image format in a block area, both schemes of Quality
Filtering (QF) and Affine Transformation Prediction (AFTP) are proposed for automatic tie point extraction and measurement with a
better and satisfactory efficiency. Tests are done by using aerial images taken with the RMK DX camera in Taiwan. Also, high
precision ground check points are adopted to evaluate the quality of the results. They show that a high density of 3D object points
are extracted and determined. Furthermore, the automatic tie point selection and measurement is done efficiently even under the
circumstance that no priori-knowledge on image overlap is available. Also, ground check points show that the accuracy of photo
coordinates is 0.21 pixels, namely it reaches a subpixel level.
1. INTRODUCTION The SGM method is very well-known in the field of computer
vision, and used for finding corresponding pixels in a pair of
One of the up-to-date issues in photogrammetry is dense images or multiple ones. It assumes the image orientation data
matching, especially pixelwise matching of aerial images. and the information of image overlapping are known. In
Matching results provided by local stereo matching methods photogrammetry, unknowns of image orientations and object
like Normalized Cross-Correlation (NCC) and Least Square coordinates can be solved by bundle block adjustment, which is
Image Matching (LSIM) are in general not reliable enough. — a primary process of geomatic data acquisition (Heipke, 1997).
Global matching (GM) of highly overlapping images increases To further increase the degree of automation of modern aerial
the reliability, but its computational complexity is too high. The triangulation and geomatic data acquisition, this paper proposes
commercial software Photosynth/Geosynth by Microsoft a new method based on SIFT for dense point cloud matching
Corporation utilized the GM technique for dense matching and without the need on any image overlap information.
stitching pictures together, with Virtual Earth, encouraging
businesses to combine the two technologies (Computerworld, 2. METHOD
2012). In order to reduce the runtime of GM, the German
Aerospace Center (Deutsches Zentrum für Luft- und Raumfahrt, 2.1 Main Processing Phases
DLR) developed the semiglobal matching (SGM) method. Both
SGM and its extensions are described in typical publications To take a more compatible architecture into account, a scale
like (Hirschmueller, 2008 and 2011). They are adopted by the and rotation invariant method is selected for automatic tie point
commercial software 3D RealityMaps to perform accurate and measurements, namely the well-known scale invariant feature
reliable dense point cloud matching, and are useful for many transform (SIFT) technique. SIFT belongs to the class of
applications like 3D reconstruction of object surfaces, feature-based matching, and includes two main processing
especially on local surfaces with occlusions, edges, fine ^ phases — keypoint extraction and keypoint matching (Lowe,
structures, and low or repetitive textures (Siegert, 2011; ^ 2004). Keypoint extraction includes Gaussian filtering and
RealityMaps, 2012). computation of DoG (Difference of Gaussian) at different
image pyramid level to detect the extreme values. Those pixels
For example, the pixelwise, Mutual Information (MI)-based with these extreme values are selected keypoints, described by
matching cost is used for compensating radiometric differences means of a descriptor defined by a 128 dimensional vector.
of input images. The method offers a very good trade off ^ Then keypoint matching is simply to calculate the Euclidean
between runtime and accuracy, particularly at object borders. ^ distances from one keypoint descriptor on the left image to
SGM has participated in several tests and evaluations. The ^ another keypoint descriptors on the right image, i.e. a pair of
Middlebury stereo pages (Scharstein and Szeliski, 2011) images at one time. If the distance ratio (the shortest Euclidean
currently list 108 stereo methods. The consistent SGM that is distance divided by the second short one) is smaller than the
modified for structured indoor scenes has a Rank of 30 and an threshold, then the keypoint is matched. Thus, the one on the
average error of 5.8%. left image is matched to another one on the right image.
Otherwise, the matching for this keypoint on the left image fails.
* Corresponding author. This is useful to know for communication with the appropriate person in cases with more than one author.
68
The
unti
The
Wit
infc
to
(abl
2).
ima
ima
of e
and
mat
con
cod
Fig
poi
resi
left
tabl
(PN
don
ima
inq
key
are
the
Lug
Bx
Ls
x
fs
Fur
esp
divi
cap
ado
