09-04-2014, 04:29 PM
An Omnidirectional Vision-Based Moving Obstacle Detection in Mobile Robot
Omnidirectional Vision-Based Moving.pdf (Size: 977.71 KB / Downloads: 22)
Abstract:
This paper presents a new moving obstacle detection method using an optical flow in
mobile robot with an omnidirectional camera. Because an omnidirectional camera consists of a
nonlinear mirror and CCD camera, the optical flow pattern in omnidirectional image is different
from the pattern in perspective camera. The geometry characteristic of an omnidirectional camera
has influence on the optical flow in omnidirectional image. When a mobile robot with an
omnidirectional camera moves, the optical flow is not only theoretically calculated in
omnidirectional image, but also investigated in omnidirectional and panoramic images. In this
paper, the panoramic image is generalized from an omnidirectional image using the geometry of
an omnidirectional camera. In particular, Focus of expansion (FOE) and focus of contraction
(FOC) vectors are defined from the estimated optical flow in omnidirectional and panoramic
images. FOE and FOC vectors are used as reference vectors for the relative evaluation of optical
flow. The moving obstacle is turned out through the relative evaluation of optical flows. The
proposed algorithm is tested in four motions of a mobile robot including straight forward, left
turn, right turn and rotation. The effectiveness of the proposed method is shown by the
experimental results.
INTRODUCTION
It is indispensable to detect the obstacles and free
space for locomotion of the mobile robot in real-world
environment. Recently, vision-based environment
detection methods have been actively developed in
robot vision [1-3]. The vision system can provide not
only a huge amount of information but also color
information in the populated environment. Currently,
because the omnidirectional vision system [4,5]
supplies a wide view of 360 degrees, they have been
popularly used in many applications such as the
motion estimation [6-10].
MOVING OBJECT DETECTION
ALGORITHM
In this section, the moving objects detection
algorithm in panoramic image is described. The
algorithm flowchart for detecting moving objects is
shown in Fig. 7. The moving objects detection
algorithm is summarized by the following procedure.
Step 1: The omnidirectional image captured from
CCD camera is monochromized to gray image.
Step 2: The omnidirectional image is expanded into
panoramic image using (2). The intensity of each
pixel in panoramic image is determined using the
linear interpolation method.
EXPERIMENT AND DISCUSSION
The omnidirectional camera is mounted on top of a
mobile wheelchair robot moving in indoor
environment. The proposed moving object detection
method was applied to four motions mode including
straight forward, left turn, right turn and rotation. The
captured omnidirectional image size is 320 × 240
pixels. The expanded panoramic image size is
630 × 87 pixels. The frame rate of input image is 15
fps (frames per second). The color of input image is
24 bit RGB color. An optical flow was computed
using a spatial local optimization [28]. The sample
frames of detected moving objects in four motions are
shown in Fig. 14. It is difficult to evaluate reasonably
the proposed method in the dynamic motion of a
mobile robot. The number of whole detected optical
flow regions have been influenced on not only the size
of spatial local region used for calculating optical flow
but also the number and velocity of moving objects
and distance between mobile robot and moving
objects.
CONCLUSION
This paper presented a new moving object detection
method in mobile robot with an omnidirectional
camera. The optical flow was estimated using the
spatial local optimization method for fast processing
time. The optical flow which has been affected by the
geometry characteristic of a hyperbolic mirror was
investigated in omnidirectional and panoramic images.
In case of translation of a mobile robot, FOE and FOC
vectors were derived from the estimated optical flow
were. FOE and FOC vectors were used as reference
vectors to detect moving objects. In case of rotation of
a mobile robot, the size and direction of optical flows
were only used to detect moving objects. The moving
obstacle was detected through the relative evaluation
of optical flows. Especially, the filtering was
suggested to eliminate the noise sensitivity of optical
flow. Results of experiments using real moving
images showed the effectiveness of the proposed
method for detecting moving objects in mobile robot.
As future work, it is necessary to reduce the vibration
of mobile robot. Particularly, if we use (14), FOE and
FOC vectors can be extracted from linear velocities Vx
and Vy of the mobile robot which can be calculated by
the motor encoder.