2011 Third International Conference on Computational Intelligence, Modelling & Simulation

Motion Detection and Analysis with Four Different Detectors

Ching Yee Yong, Rubita Sudirman Kim Mey Chew

Faculty of Electrical Engineering Faculty of Computer Science and Information System
Universiti Teknologi Malaysia Universiti Teknologi Malaysia
Johor, Malaysia Johor, Malaysia
[email protected]/[email protected] [email protected]

Abstract— A change in measurement of speed or vector of an environment and thorough discussion on the developmental
object or objects in the field of view is called motion. Detection tool or assessment and analysis on various motions will be
of motion can be achieved by electronic devices or mechanical explained in section 4. Finally, the last section contains the
devices that interact or quantify the changes in the given conclusions, future developments and possible enhancement
environment. In this paper, a monitoring system is being and improvement on this study.
enhanced utilizes motion detection technology with modified
internationally recognized algorithms, implemented in C sharp II. PROBLEM FORMULATION
and Matlab programming language. The result of this study is
expected to be beneficial and able to assist users on effective The hypothesized function of this monitoring system is to
motion detection and analysis. Four different motion detectors provide the users with information about the detection of an
are being compared. The assessment includes three trials in object’s movement in order to deliver useful information
three different speeds of motion and Morph filter has given a about the motion with connection to the theory of motion
better and smooth detection. In conclusion, an effective motion through movement. The subjective experience of motion is
assessment and monitoring system has been developed for the simple hypothesized as the feedback from tiny movement
improvement of the motion detection ability. and differences between two frames. In this study, the
practice item of motion was focused on comparison among
Keywords- Motion; motion detection; morph; motion
detectors
four detectors.
The objectives of the fundamental studies are as follows:
 To implement the algorithm for motion detection analysis
I. INTRODUCTION
in a newly developed monitoring system.
Motion can be detected by: sound (acoustic sensors),
 To compare and analysis among four detectors used
opacity (optical and infrared sensors and video image
(Current and Previous, Pixellate, Blob Counter and
processors), geomagnetism (magnetic sensors,
magnetometers), reflection of transmitted energy (infrared Morph).
laser radar, ultrasonic sensors, and microwave radar sensors), Counterbalancing should be instituted to control for order
electromagnetic induction (inductive-loop detectors) and effects and efforts made to ensure that all subjects complete
vibration (triboelectric, seismic, and inertia-switch sensors) objectives. New task need to be developed drawing on the
[1]. best features of existing tasks.
There are many approaches for motion detection in a A wide variety of human-machine interactions learning
continuous video stream. All of them are based on techniques have been used in human monitoring approaches
comparing of the current video frame with one from the for motion analysis. This employs a large number of visual
previous frames or with something that called as and physiological features, a fact which usually impedes the
background. In this paper, there are four approaches are used training process [2].
and comparison is made to find out a best detector for an In this paper, an effective monitoring system for motion
effective motion detection. assessment is presented. Several aspects for example the
The system supports the AVI files, JPEG from internet speed of the subject are considered while the assessment is
cameras, MJPEG (motion JPEG) stream, local capture being done. This monitoring system does not only serves
device and webcam. motion movement as assessment, but also important in
This paper is divided into six sections. The first section managing the subject data effectively and providing
mainly introduces the whole study. It provides the general scientific information about object’s movement.
overview of the motion and analysis system. The second
section includes the objectives of this study, which describes
the aims that needed to be achieved. The third section
discusses the background studies, literature review and the
study implementation. A specification list of the computer

978-0-7695-4562-2/11 $26.00 © 2011 IEEE 46

DOI 10.1109/CIMSim.2011.18
uthorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on September 08,2023 at 13:10:45 UTC from IEEE Xplore. Restrictions apply
 III. LITERATURE REVIEW C. Pixellate Filter
Pixel art scaling algorithms are image scaling algorithms
A. Current Frame vs. Previous Frame specifically designed to upsample (enlarge) low-resolution
One of the common approaches for a motion detector is pixel and line art that contains thin lines, solid areas of
to compare the current frame of a streaming video with the colour rather than gradient fills or shading and has not been
previous frame. It is very useful in video compression anti-aliased [7].
especially in estimation of changes, writing only the It is a simple pixellation algorithm that takes the median
changes and not the whole frame. of the saturation over each square and the mean of the
Firstly, Difference and Threshold filters are used to luminescence. The main idea of this filter is to replace
distinguish the difference regions between an original gray pixels in some rectangular area with their average value and
scaled frame and the previous video gray scaled frame. An produces a nicely graduated luminescence by taking the
image with white pixel on the difference regions is obtained mean of the hue is prone to error since adjacent hue indices
on the specified threshold value. A motion event can be can be radically different colours [8].
signaled if the value is greater than a predefined alarm level Hence, the pixellated versions of the current and
[3]. background frames are produced. The next step is to move
Then, erosion filter is used to remove random noisy the background frame towards the current frames and the rest
pixels since mostly of cameras produce a noisy image. The is only the main processing step [9].
Erosion filter is a morphological filter that changes the
D. Blob Counter
shape of objects in an image by eroding (reducing) the
boundaries of bright objects and enlarging the boundaries of The idea of blob detection is to turn an image of pixels
dark ones. It is often used to reduce or eliminate small into a collection of “super pixels”. Blob detection works by
bright objects [4]. grouping sets of pixels that fall within a particular range of
This filter assigns minimum value of surrounding pixels colours. The rest of the image can then be discarded and the
to each pixel of the result image. Surrounding pixels, which resulting superpixels can either be highlighted on the image
should be processed, are specified by structuring element: 1 or sent for further processing. This creates very specific
to process the neighbour or -1 to skip it. It is very useful for areas of interest, which can be processed further, as each
binary image processing, where it removes pixels, which are region has a set of coordinates.
not surrounded by specified amount of neighbours. It gives This approach has a great possibility for performance
ability to remove noisy pixels or shrink objects [5]. optimization. It provides the number of objects, position and
At this stage, an actual motion is obtained since mostly dimension on a binary image. Blob counter is a very useful
only the interest regions are being detected. From the below feature and can be applied in many different applications. It
picture, the disadvantages of the approach had been can count objects on a binary image and extract them. The
discovered. If the object is moving smoothly, small changes idea comes from “Connected components labeling,” a filter
from frame to frame was received. Hence, it is a problem and that colors each separate object with a different color [10].
difficulty to get the whole moving object. Things become Blob extraction is also very useful because once the super
worse when the object is moving so slowly and the pixels have been identified, a simple calculation can be
algorithms will not give any result at all. performed to see if the blob meets per determined conditions,
such as height, width or shape [11].
B. Current Frame vs. First Frame
Another approach is to compare the current frame with IV. METHODOLOGIES
the first frame in the video sequence. Comparison between In this section, the procedures and the methods used for
these two frames were resulting the whole moving object the monitoring system will be described. C Sharp and Matlab
independently of its motion speed. software are used in all the programming section. Generally,
At the beginning, the first frame of video sequence is set the flow of the motion monitoring system is as in Figure 1.
as background frame. The upcoming frames are always First, a short motion of hand waving with a ring on the
comparing with this background frame. The most important last finger as reference is recorded real-time with a laptop
approach here is to “move” the background frame slightly in webcam. The data is then fed into system for analysis using
the direction of the current frame on the specified amount four different detectors. Result of motion detection is
for example used one level per frame. Then, the colours of generated parallel with the data input. All the detected
pixels in the background frame are changed by one level per motions are highlighted with red colour.
frame [6]. Motion alarm is added into the motion detection system
The most efficient algorithms in detection motion are by calculate the amount of white pixels on the difference
building the scene or background frame as a reference to image between current frame and background frame.
make comparison with the current frame. Some approaches
listed below are rather simple if compared with the most
which are too complex.

47

uthorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on September 08,2023 at 13:10:45 UTC from IEEE Xplore. Restrictions apply
 transition. Most often it is used to depict one person turning
into another through technological means or as part of a
E-Motion fantasy or surreal sequence. Traditionally such a depiction
would be achieved through cross-fading techniques on film.
Current vs. Since the early 1990s, this has been replaced by computer
Previous Frame software to create more realistic transitions [12].
The idea of the filter is to preserve specified percentage
Pixellate of the source filter and to add missing percentage from
Data input Result
overlay image. So, if the filter was applied to source image
Blob Counter with percent value equal to 60%, then the result image will
contain 60% of source image and 40% of overlay image.
Morph Applying the filter with percent values around 90% makes
background image changing continuously to current frame.
This new filter has two benefits that it is much simpler
Figure 1. Block diagram for motion assessment and training monitoring to understand and produce better performance due to the
system. efficiency of implementation of filter.

A. Proposed Approach
Morphing is a special effect in 2D images that changes
(or morphs) one image into another through a seamless
V. RESULT

Figure 2. Result image (from left) using Current vs. Previous Frame, Pixellate Filter, Blob Counter and Morph Filter with increasing motion speed.

48

uthorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on September 08,2023 at 13:10:45 UTC from IEEE Xplore. Restrictions apply
 Figure 3. Result image (from left) using Current vs. Previous Frame, Pixellate Filter, Blob Counter and Morph Filter for a small table tennis ball detection.

VI. DISCUSSION The monitoring system is designed around four essential

The advantage of this monitoring system over other is the qualities: validity, reliability, impact and practicality.
ability of this monitoring system to provide an effective and Validity is normally taken to be extent to which an
easy method for user. It is important to consider the assessment can be shown to produce scores which are an
assessment in all the aspects including speed and accurate reflection of a human’s motion true level.
background. Reliability concerns the extent to which assessment results
The most appropriate monitoring available is this system are stable, consistent and accurate, and therefore the extent
able to monitor a progress in a long time. Result is presented to which they can be depended on for making decisions
in Figure 2. Four different detectors are used in the system about the motion. Impact concerns the effects, beneficial or
starts from left is comparing current frame with previous, otherwise, which an examination has on the motion.
pixellate filter, blob counter and morph. All these detectors Practically can be defined as the extent to which an
are working smoothly with high efficiency during slow assessment is practicable in terms of the resources needed to
motion about 6.40 frames per second (fps). produce and administer it.
On the other hand, when the system is applied on very There is a lot to do with this motion detection monitoring
fast motion around 36 fps, morph filter is able to detect the system depending on the imagination. One but not the only
object’s motion clearly with a tiny lagging rather than the one straight forward application for this system is video
previous three detectors but there is an interesting finding surveillance. It also can be applied on the incredible thing
that pixellate detector is unable to detect a very fast likes gesture recognition, behavioral analysis and gait
movement. It will automatically decrease the motion to analysis.
become slow motion for an analysis due to its characteristic VII. CONCLUSION
of enlarge low resolution image and it takes longer time to
process the mean for detection. Some sections of the trial In conclusion, an effective monitoring system for motion
motion like subject’s palm center and subject’s ring on the detection and assessment tool has been developed. The level
of motion is used as the input for the monitoring system to
last finger were successfully detected by the last detector.
generate assessment to the motion detection to the particular
Another motion video was analyzed using this four
person specifically. Results and findings show that the
detectors and the result showed that morph filter able to monitoring system is effective and consistent in producing
detect the fast movement of table tennis ball effectively for relevant results to the detected motion. This monitoring
a long period. The detector can detect, recognize and coping system can be developed in the algorithm for speech
out a zone of interest of every moving object in a frame. recognition system in order to obtain more accurate and
The result indicates that morph filter is better in motion reliable voice input.
detection for the monitoring system. The detection is In future, this system will be upgraded to mobilize
running smoothly and available in fast motion detection. resources to provide the necessary infrastructure, supplies
This monitoring system also could be revised to make and materials needed to ensure every assessment is
them more user-friendly, with a focus on issues such as achieving the motion analysis potential. This is important to
layout, illustrations, message, information, and cultural increase the reliability and effectiveness of this monitoring
appropriateness. It provides full functionality for the entire system.
assessment cycle: authoring, scheduling, administering and A more detailed concept of motion detection will be
rating. A non-technical rationale for using the procedure, more useful in later processing stages. As in image flow
including the type of information and what we can do with algorithm, all the information is need to be incorporated on
this information will be provided. It is a premier and the direction of motion. Optimization in realization is very
affordable personal computer-based assessment system for important for a optimize solution from the beginning.
academia, government and business users.

49

uthorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on September 08,2023 at 13:10:45 UTC from IEEE Xplore. Restrictions apply
 ACKNOWLEDGMENT [5] Ramprasad, P. and Randal, C. N. (1994). “Recognition of Activities”,
Proc. International Conference on Pattern Recognition, Jerusulem,
The authors are deeply indebted and would like to Israel, A815-820.
express our gratitude to the Universiti Teknologi Malaysia [6] Tetsuya, M., Makoto, N., Tomohiro, Y. and Shinji, T. (2010).
and Ministry of Higher Education (MOHE) for supporting “Comparison of Color Space in Extraction of a Hand Region
and funding this study under Research University Grant for Computer Human Interface Using Color Image
Processing”, Technical report of IEICE. PRMU 98(528).
(QJ13000.7123.00H44) and Zamalah Scholarship Scheme.
Our appreciation also goes to the Electronics and Biomedical [7] Thomas, K. (1999). “Fast Blit Strategies: A Mac Programmer's
Guide”. MacTech.
Instrumentation (bMIE) for their cooperation in the research
[8] “Eagle (idea)”. Everything2. 2007-01-18.
work.
http://thread.gmane.org/gmane.linux.redhat.fedora.legal/661
REFERENCES [9] Stepin, M. (2007). “hq3x Magnification Filter”.
[1] Randal, C. N. (1991). “Qualitative Detection of Motion by a Moving Retrieved 2007-07-03.
Observer”, Proc. IEEE Conference on Computer Vision and Pattern [10] AForge.NET. (2009, 03 27). BlobCounter.cs.
Recognition, Maui Hawaii, 173-178. Retrieved 03 01, 2010, from code.google.com:
[2] Loui et al. (1990). “High-Speed Architectures for Morphological http://code.google.com/p/aforge/source/browse/branches/1.7/Sources/
Image Processing”, Nonlinear Image Processing, pp. 145-156. Imaging/BlobCounter.cs
[3] Sternberg, S. R. (1983). IEEE Computer, Biomedical Image [11] Amit, Y. (2002). 2D Object Detection and Recognition: Models,
Processing, CytoSystems Corporation. Algorithms, and Networks. Chicago: Cambridge, Mass. MIT Press,
[4] Heijmans, H. J. and Ronse, C. (1990). The algebraic basis of 2002.
mathematical morphological-I: Dilations and erosions, J. Comput [12] Kalman, R. E. (1960). A new approach to linear filtering and
Vision, Graphic, Image Process, Vol. 50. prediction problems, Transactions of the ASME -- Journal of Basic
Engineering, Series D, pp. 35-45.

50

uthorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on September 08,2023 at 13:10:45 UTC from IEEE Xplore. Restrictions apply