![](bg1.png)
APPEARANCE-BASED AND ACTIVE 3D OBJECT RECOGNITION
USING VISION
F. Trujillo-Romero and M. Devy
CNRS; LAAS; 7 avenue du Colonel Roche, 31077, Toulouse, France
Universit
´
e de Toulouse; UPS, INSA, INPT, ISAE; LAAS; Toulouse, France
Keywords:
Recognition, Appearance-based, Active strategy, Robotics.
Abstract:
This paper concerns 3D object recognition from vision. In our robotics context,an object must be recognized
and localized in order to be grasped by a mobile robot equipped with a manipulator arm: several cameras
are mounted on this robot, on a static mast or on the wrist of the arm. The use of such a robot for object
recognition, makes possible active strategies for object recognition. This system must be able to place the
sensor in different positions around the object in order to learn discriminant features on every object to be
recognized in a first step, and then to recognize these objects before a grasping task. Our method exploits
the Mutual Information to actively acquire visual data until the recognition, like it was proposed in works
presented in (Denzler and Brown, 2000) and (Denzler et al., 2001): color histogram, shape context, shape
signature, Harris or Sift points descriptors are learnt from different viewpoint around every object in order to
make the system more robust and efficient.
1 INTRODUCTION
Object recognition is a task that a human being car-
ries out in an instinctive way. Many factors make
difficult such a task: illumination conditions, relative
camera-object positions, occlusions, etc. So, endow-
ing a robot of this capability is not easy.
Many researchers in Computer Vision have
worked in this topic, providing many publications.
During the last decade, many improvements have
been provided by the appearance-based methods.
Lowe et al. (Lowe, 1999; Lowe, 2001) propose to ex-
ploit points extracted from images because their pho-
tometric properties are invariant with respect to small
camera motions: such points are extracted by Differ-
ences of Gaussians (DOG) or other scale-invariant de-
tectors (e.g. the Scaled Salient Patches of Kadir), and
then are characterized by a descriptor: the SIFT one
has been proven to be the more discriminant. Hebert
et al. (Johnson and Hebert, 1996) (Zhang and Hebert,
1996) have developed an approach for object recogni-
tion, using Spin Images, i.e. a map of images acquired
when a camera is moved around an oriented point.
Fergus et al. (Fergus et al., 2003) and Ke et al. (Ke
and Sukthankar, 2004) have proposed independantly
PCA methods in order to improve the original Lowe
approach based on SIFT descriptors.
In the Computer Vision community, the typical
strategy consists in exploiting only one image in or-
der to recognize an object. Using robots to move sen-
sors, allows active recognition methods, since the sys-
tem can place the sensor in the scene in function of
the current status of the recognition process, i.e. of
what has been perceived and understood from previ-
ous images. In (Trujillo-Romero et al., 2004), we pro-
posed an active recognition method based on the mu-
tual information, by exploiting only color attributes of
the analyzed objects. In (Jonquires, 2000), we pro-
posed another active strategy for the recognition and
the localization of polyhedral objects from a camera
mounted on the wrist of a manipulator: Bayesian Be-
lief Networks (BBN) was built during a preliminary
step, to learn (1) how to select the best strategies along
the recognition step, (2) the best perceptual groupings
to provide hypothesis from an initial image and (3) the
best camera positions to verify these hypothesis from
next images.
Our problem concerns the recognition and local-
ization of objects to be grasped by a robot: objects
must be recognized by the system shownoin figure 1;
learning and recognition functions must be performed
on line, in a human environment, typically at home,
where illumination conditions cannot be controlled.
The illumination variability makes non efficient ap-
417
Trujillo-Romero F. and Devy M. (2009).
APPEARANCE-BASED AND ACTIVE 3D OBJECT RECOGNITION USING VISION.
In Proceedings of the Fourth International Conference on Computer Vision Theory and Applications, pages 417-424
DOI: 10.5220/0001805604170424
Copyright
c
SciTePress