Using Image Processing as a Measuring Device in Close Loop Control System and System Behavior Analysis

Full text

Using Image Processing as a Measuring Device in Close
Loop Control System and System Behavior Analysis
MAJID ALITAVOLI
SALAR BASIRI
AHMAD BAGHERI
Mechanical Eng. Department
Guilan University, Rasht
IRAN
tavoli@guilan.ac.ir, salarbasiri@noavar.com, bagheri@guilan.ac.ir
Abstract: In this paper, the common principles of digital image processing are investigated.
By using these principles, one can deploy simple devices such as a PC or webcam to establish
a system based on Image processing techniques in order to test different algorithms. The use
of color software filters increases the need for processing power of image processor, but
based on the ease of a capability and inexpensiveness of modern computers can reduce the
total system expenses. Further in this research, several image processing systems in which
cameras are used as feedback sensor in close loop control systems. These cameras can aid in
obtaining kinematic parameters such as speed and position. In some cases the use of this
method, due to its system physical status can be the best method. In some other situation this
method could be used alongside of other choices. The chief advantage of this technique is that
the instrumentation tools can not affect the case under study.
Key-Words: Image processing, Vision feedback, Tracking systems, Adaptive control, pattern
recognition
1 Introduction
Nowadays, image processing has found various
applications in industry and in some cases it can
be used as the only data collecting technique.
A variety of image processing has been
developed so far such as human motion capture
and analysis[1], tracking system[2], human-
computer interaction system[3], and etc.
nearly all of them employ a camera for acquiring
data which is used to control several actuators.
Also be used in extraction of necessary
parameters and use the picture frames saved in
certain processing action.
The main component of an image processing
system is shown in Figure1.
Fig1. The elements of an image processing system
Transfer function of each component can be
modeled analytically, determined experimentally,
or taken from manufacturers' specifications.
The lenses, for example, can be assumed
diffraction limited. The computer operation may
or may not be linear, but this is the only
subsystem in fureig1 that is directly under the
users' control [4].
In simple systems a webcam can be used as
sensing unit for image capturing. These cameras
are of CMOS and CCD type. CCD sensors show
a higher quality image than other type. Images
can be processed by Digital Signal Processor
(DSP) or personal computer (PC). The choice
depends on the availability and the needed
processing speed.
In this paper some of the image processing
system which were developed by the authors will
be described.
The first system, by using image processing
techniques kinematic parameters of a passive
bipedal walking is determined. In the second
system, a puzzle solver robot which employs
machine vision for positioning robot TCP is
described. Also the numerical pattern recognition
Proceedings of the 6th WSEAS International Conference on Applications of Electrical Engineering, Istanbul, Turkey, May 27-29, 2007 15

is explained in this section. The third system
introduces a laser tracking system in which a
camera is used as a sensor for recognition of
target position.
The fourth and the last case describe a prototype
test crane which has employed a machine vision
system to determine free load status in mid air.
This system provided one of the most accurate
techniques in position finding of material.
2 Image Matrixes
Image capture sensor is formed from some light
sensitive elements. The voltage of each element
is proportional to the intensity of radiation light.
The image forms on the sensor plane by variating
light propagation tools such as convergent lenses
(Figure2).
Fig2. Image formation
The voltage of each image plane element is
converted to a binary code by an ADC and then
transferred to main processor by a
communication cable.
The communication protocol is chosen from
different S-video, IEEE1394, RS170, or USB
type. Based on their application, each of them
have their own advantages or drawbacks.
In case of the use of a PC as processor equipped
with Microsoft windows as OS, using Video For
Windows (VFW) programming technique is
recomended. This technique makes a simple
communication with all image capturing devices.
At the end, the expected images as RGB or
CMYK format and as a 3D matrix which its
elements are image pixels color value will be
formed.
3. Color separation filters and noise
effect reduction
Separating a particular color or color node is
common need in various color image processing
systems. In most cases in order to separate and
pass a particular color colored filters are used.
A less expensive and more flexible solution is the
utilization of software filters specially when
using webcam as image capturing sensor.
Different digital image saving methods are
developed among them, RGB color space have
more application in color image processing.
In this system the search space is determined by
R, G, and B axis. Each of the amounts of R, G
and B for a singular pixel can be in the range of 0
and 255. For separating of a particular color the
amount of B, G, and R is compared to a desired
value. For example to identify the yellow color
the amount of the R and G must be more than
200 and amount of B must be less than 150.
The color space of these Conditions is shown in
Figure3.
Fig6. Rectangular color space
After separation of desired colored nodes, for
noise effect reduction a low pass filter can be
used. Matrix coefficients for such a filter are
calculated as a moving average filter as follows:
=
⎥
⎥
⎥
⎦
⎤
⎢
⎢
⎢
⎣
⎡
919191
919191
919191
LPF Mask
A After separation of needed pixels and elimination
of noise, the necessary recognition algorithm is
used. This algorithm can be used to easily
determine the pixel coordinates or can be
employed to deffrentiate a model of pixel
sequences. Further, some different projects are
described.
Proceedings of the 6th WSEAS International Conference on Applications of Electrical Engineering, Istanbul, Turkey, May 27-29, 2007 16

4 Determination of Kinematic
Parameters of a Passive Bipedal
Walking Robot
In this project, the kinematic parameters of a
bipedal robot were determined. These parameters
were also obtained trough simulation techniques.
Then, the results were compared with
experimental results[5].
The physical characteristics of the fabricated
biped are shown in Figure3. This robot is
designed and manufactured based on the McGeer
Biped which was first designed and
manufactured in Cornell University in the USA
[6], [7].
Fig.3: Passive Biped Walker of the
University of Guilan
Figure 3 shows the schematic of the bipedal robot
The plane coordinate is divided to 3 geometrical
zones and also 2 color zones. All zones should be
distinguished from each other. The geometrical and
color zones, is shown in Figure 3.
The amounts of the position areas are given to the
image process program. According to the separating
method of the position areas and colored areas, it is
assured that the obtained point is unique.
As the relation of the distance between the planes -
where the nodes are inside them – and the mean
distance of them to camera is very close , all the
nodes can be assumed in the same plane. It is due to
the effect of immeasurable factors on the transfer
function of the pixel positions to the real positions.
Nodes position accuracy depends on camera
resolution and distance between image plane (CCD
array) and object plane. Nodes appear like color
area in image and center of that area is node
position.
Fig.8: Geometric and color zone
In cases where the mid-legs is covered by the leg
that is faced to the camera - which prevents it to
ind the nodes- Linear Regression method is used to
obtain closest datum to the unknown data. For
obtaining angular velocity and acceleration the
following formulation (5 points differential) is used.
To write the program, Delphi 7.0 is used. The GUI
program is shown in Figure 9.
Fig.11: The Program GUI designed by Delphi 7.0
5 The Puzzle Solver Robot
The robot introduced in this section is designed
to solve numeral puzzles. Its control is totally by
computer and has vision capability which leads
its arm to certain places. In the control loop of the
arm for identifying the top of the arm, image
processing technique is used. Because of its polar
movement on the plane, the design of the robot is
considered an optimized one. The control orders
are given to robot by a parallel port which obtains
image from a camera through USB port. For
numerical recognition a simple algorithm with
the capability of learning new patterns is used.
Noting that different conditions have not been
given to the robot, therefore the robot is highly
considered intelligent. Some of the robot
characteristics include recognizing numbers by
camera, 3 degree of freedom, polar movement of
the plane, locating the arm of robot with the use
Proceedings of the 6th WSEAS International Conference on Applications of Electrical Engineering, Istanbul, Turkey, May 27-29, 2007 17

of camera, and capability of learning simple
numerical pattern.
Such a system can be helpful as a benchmark in
testing and performing image processing
algorithms. An arm with 3 degrees of freedom as
effector and a webcam as vision sensor has been
used in the system. Processing of images should
be done using a personal computer [8].
Fig6. Puzzle Solver Robot Schematic
The robot's control program is written by Delphi7
with a total of 1600 lines of programs.
5.1 Number recognition
After distinguishing the number from the
background, the next step is to recognize them.
A novel algorithm is used to recognize numbers
using Figure 9 (this is 3 in Persian form), this
algorithm is described.
Fig9. Number Recognition Method
As observed in fig.6 the numbers of lines forming
a number in different coordinate positions in the
number plane are different. This is the key for
number recognition in this algorithm. For
example for number 3, the number of upper limes
are 3, one lines on the bottom, one line on the left
and one line on the right is used.
5.2 Controller
In figure11 a schematic diagram of
image processing based close loop controller is
illustrated.
Fig11. Image processing based close loop
controller
After number recognition and puzzle solving, all
the action that the robot must do in order to
organize the puzzle will be determined.
Based on the coordinate of each numbers in the
system, this procedure is prepared in the form of
a G-code like commands. So, the controller loop
based on coordinates will change the position of
cell. The plane is swiped once to finalize TCP.
Due to the different between the TCP color and
its background, its position will be determined.
Now the robot is ready to optimize the
movement by considering the TCP point and the
intended position. To do this, Firstly radius and
then angle of the arm will increase or decrease to
become equal to the optimum values of radius
and angle.
6 Tracking System
Image processing based tracking systems
equipped with powerful computing systems have
gained vast applications nowadays. In this
section, an image processing based tracking
system and its algorithms is described.
Tracking moving objects over time is a complex
problem in computer vision and has been an
important research subject over the last few years
[9], [10], [11]. Impressive tracking systems have
been developed for some specific applications
[12], [13]. In case of selecting one object among
several, the subject becomes more complex,
especially if the targets are of nearly same color.
In this case, system can track the target more
Proceedings of the 6th WSEAS International Conference on Applications of Electrical Engineering, Istanbul, Turkey, May 27-29, 2007 18

accurately by limiting the search window while
zoomed around the target.
In the following project, our goal is tracking of
moving target over a plane and turning it on by
laser pointer.
This device can be used as image processing
algorithm benchmark in laboratories. Figure 3
shows the image of the device.
Fig3. Image processing based laser
tracking system
6.1 Mechanical and electronic sections
For moving the laser point over objects plane,
two stepper motors are used. These motors are
connected to a gear by a worm gear. The
movement of these gears changes the position of
laser point. The necessary control command is
given by computer through parallel port to the
interface circuit. Power transistor is used in the
interface circuit. These transistors can turn off/on
stepper motor inductors due to command
received directly from parallel port. The interface
circuit can also turn the laser pointer off or on.
A web camera with 352x288 pixel resolution is
mounted on the device as well. It transfers the
images through a USB link to the computer.
6.2 Determining the necessary control commands
for laser pointer movement
After finding the target coordinates and the light
of laser pointer, the correct command must be
given to the stepper motors trough the parallel
port in order to move laser pointer to correct
direction.
In order to find the correct movement, the nearest
neighbor method is used. According to this
method, since the pointer movement in 2D image
plane is limited to the eight boxes around the
pointer position. To find the right direction, it is
enough to find the distance of each of the eight
boxes from the target. Then, the proper direction
is toward the box which has the less distance to
the target (Figure 4).
Fig 4. nearest neighbor method
7 Position recognition of free load of
a crane
The goal in this project was to find the position
of a free load in mid air. Since the load is hanged
by a flexible rope, using angles sensor can not be
accurate. Hence, the use of a camera can provide
the simplest technique for this purpose.
Fig7. Crane and free load Schematic
In figure 5 the schematic of the system is shown.
Similar to tracking systems, in this system the
windowing technique has been used for reducing
the time of image processing.
8 Results and Discussions
In this paper several instrumentation systems for
analysis of different behaviors which employed
vision techniques, were presented. In many
situations which we need to determine kinematic
Proceedings of the 6th WSEAS International Conference on Applications of Electrical Engineering, Istanbul, Turkey, May 27-29, 2007 19

parameters of a system, specially position
parameter, image processing can provide one of
the best alternative. Improvement methods for
noise elimination and pattern recognition in
which the increase of processing power for
computing systems is presented, can provide
powerful and efficient tools for researchers in
various field of science and technology. The use
of these techniques not only results in achieving
better results but will help to save time and
money in total.
References:
[1] J.K. Aggarwal, Q. Cai, "Human motion
analysis: a review", ComputerVision and Image
Understanding 73 (3) (1999) 428–440.
[2] M. Alitavoli ,S. Basiri, S. Basiri, " Image
Processing Based Tracking System ", WSEAS
TRANSACTIONS on SIGNAL PROCESSING, Issue
12, Volume 2, December 2006, pp.1558-1662
[3] Robert Ward, " An analysis of facial movement
tracking in ordinary human–computer interaction"
Interacting with Computers 16 (2004) 879–896
[4] Kenneth R. Castleman, Digital Image
Processing, Prentice Hall, New Jersey, 1996
[5] A. Bagheri ,A. Hajiloo ,S. Basiri
"Determination of Kinematic Parameters of a
Passive Bipedal Walking Robot Moving on a
Declined Surface by Image Processing",
WSEAS Transaction on Computer, Vol4,
Nov2005, pp1718-1724
[6] M, Garcia, “Stability, Scaling and Chaos in
Passive-Dynamic Gait Models”, A Dissertation
Presented to the Faculty of the Graduated School
of Cornell University, January 1999.
[7] A. Hajiloo, " Design and Manufacturing of a
Passive Biped Walker" , B.Sc. Dissertation,
University of Guilan, 2004
[8] M. Alitavoli, S. Basiri, H. Mallaei, S.
Rezazade osmanvandani “Application of Image
Processing For Solving Numerical Puzzles Using
A 3 DOF Robot”, WSEAS Transaction on
Circuits and Systems, Vol 5, Sep2005, pp1452-
1458
[9] Lowe, D.G., “Robust model-based motion
tracking through the integration of search and
estimation,” International Journal of Computer
Vision, vol.8:2, pp. 113-122, 1992.
[10] Coombs, D., and Brown, C., “Real-time
smooth pursuit tracking for a moving binocular
robot,” Proc. IEEE, pp.23-28, 1992.
[11] Huttenlocher, D.P., Noh, J.J., and Rucklidge,
W.J., “Tracking non-rigid objects in complex
scenes,” Proc. IEEE, pp. 93-101, 1993.
[12] Dickmanns, E.D., Graefe, V., “Applications
of dynamic monocular machine vision,” Machine
Vision and Applications, pp. 241-261, vol. 1,
1988.
[13] Frau, J., Casas, S., Balcells, Ll., “A
dedicated pipeline processor for target tracking
applications,” Proc. IEEE International
Conference on Robotics and Automation, pp.
599-604, 1992.
Proceedings of the 6th WSEAS International Conference on Applications of Electrical Engineering, Istanbul, Turkey, May 27-29, 2007 20