Вы находитесь на странице: 1из 4

8th IEEE International Conference Humanoid, Nanotechnology, Information Technology

Communication and Control, Environment and Management (HNICEM)


The Institute of Electrical and Electronics Engineers Inc. (IEEE) Philippine Section
9-12 December 2015 Water Front Hotel, Cebu, Philippines

Real-Time Integrated CCTV Using Face and


Pedestrian Detection Image Processing Algorithm For
Automatic Traffic Light Transitions
Cyrel O.Manlises, Jesus M. Martinez Jr. , Jackson L. Belenzo,Czarleine K. Perez,Maria Khristina Theresa A. Postrero
Mapua Institute of Technology
School of Electrical, Electronics and Computer Engineering
Manila, Philippines

AbstractWith the technology today, detection of pedestrian was The application used is real-time which does not record the
introduced to decrease the accidents in the roads. It was amount of frames of images to analyze. For cross-line
implemented on the traffic lights and cars. In this paper, roadways, multi-camera surveillance is implemented to utilize
integration of face detection and pedestrian takes place on the a mathematical approach for switching emergency events. On
device called Raspberry Pi. OpenCV was used for the detections evening rush hours, surveillance processing for monitoring
and programmed in Python. The webcam was used to capture vehicle movements is essential. This study of Haar-like
images in real-time. The images was processed at the feature, for object and digital image recognition, leads to Haar
microprocessor. When pedestrians were detected, the wavelet processing which is used in real-time face detector.
microprocessor played the recorded voice and outputs it to the
The Viola-Jones algorithm is known for image and motion
connecting speaker to notify those pedestrians or the pedestrian
to wait before crossing until such time. The microprocessor sends
detection which is until now undergoing improvement for
a signal to the Gizduino to process the time of the red lights on innovation. This algorithm is also used for pedestrian
state. The microcontroller was programmed with the traffic detection. A similar related study is for determining walking
signals. When it receives a signal from the microprocessor, it pedestrians, with high accuracy detection algorithm of 90%
added an extra seconds to the red lights on state. Afterwards, from the test conducted. CCTVs of today are more advanced
when it returned, it changed back to its original on state. The than before. Pedestrian crossings are one of the places where
methods used for analyzing the data obtained were the CCTVs are installed. In some pedestrian crossings, push
Classification method and the Chi-square statistical test analysis. button devices are used instead of CCTV cameras. The
After testing, gathering data and computing each classification existing push button or Ped call button, which DeVoe and
with the use of chi-square formula, the result of the success rate Wall called in their paper, in pedestrian crossings are used by
was 87.33% that showed that face and pedestrian detection that pedestrian where the traffic light control systems time
was integrated in the system successfully worked. changes once the pedestrian pushes the button.
Index Terms-OpenCV, Raspberry Pi, Python, Gizduino The existing push button system is prone to wear, tear, and
vandalism. The absence of initiating Ped call button while
I. INTRODUCTION waiting by uninformed people was observed. Existing face
detection and pedestrian detection algorithms were not
Pedestrian push button traffic control or the Ped call button
previously used in the pedestrian detections. The proposed
that exist in most crossing areas are used to change traffic light
system, should integrate an image processing system into a
transitions by the pedestrians in order to have their turn to
CCTV camera system thus eliminating the push button
cross. Another device that is found in some areas is the
system.
Closed-circuit television cameras. They are also used to
monitor areas or things that could to help security personnels The main objective of this study was to design a pedestrian
investigate acts harmful to human, and infrastructure. Hence, detection system using image processing by integrating it into
CCTV cameras are now equipped with pedestrian detection an existing CCTV camera commonly installed on street poles.
feature to detect the presence of people. This type is used in Specifically, it aimed:
the streets to know if the pedestrian wants to cross the street,
which is similar with a Ped call system. This makes a CCTV To develop and test a face detection algorithm for
on pedestrian lane have additional functionality to help the identifying the presence of human faces and its
pedestrians consider their turn to cross. orientation.

978-1-5090-0360-0/15/$31.00 2015 IEEE


8th IEEE International Conference Humanoid, Nanotechnology, Information Technology
Communication and Control, Environment and Management (HNICEM)
The Institute of Electrical and Electronics Engineers Inc. (IEEE) Philippine Section
9-12 December 2015 Water Front Hotel, Cebu, Philippines
To develop and test a pedestrian detection algorithm
for detecting the presence of pedestrian at its proper
location in the opposite side of the pedestrian crossing,
and;
To integrate and test the algorithms into a simulated
pedestrian crossing and traffic light system.
The study did cover small-scale roadways or two lane road
path. Also, the system did not cover security purposes such as
identifying face and record video for monitoring accidents and
crimes. The study was tested on daylight and the weather was
neither rainy nor had snow.
Face detection for pedestrians was helpful to the
community providing a method of traffic control, without the
presence of traffic enforcers for either gestural manipulation Fig. 1. Framework of the System
or button manipulation. This was to eliminate the Ped call
button traffic control system. It would be feasible when the
system was integrated with CCTV cameras and with this
system.
II. METHODOLOGY
The system was composed of webcam, Raspberry Pi,
Gizduino for traffic signal and recorded voice notifier using a
portable speaker. The webcam was connected to the raspberry
pi where the inputs of the images were processed. The
raspberry pi analyzed the image. The program in the Raspberry
Pi would detect the front face of the pedestrian. If detection of
face was found, a recorded voice would play using the speaker
and the traffic lights red light would be longer.
A. System Flow
Images captured were analyzed using the pedestrian
detection followed by the face detection if there was a
pedestrian found in the images analyzed. If there was a face
found, it played the recorded voice and sent a signal to the
Gizduino, which was programmed with the traffic light system.
The red light in the traffic light would be set to 45 seconds
when the raspberry pi detected a face. If none was found, it
would be 30 seconds.
B. Software Coding
The two detections were integrated using Python language.
The traffic transition was in C language.
C. Set - Up
The webcam was faced in the front of the end side of the
pedestrian lane with respect to the length of appropriate vision
by the camera to the pedestrians. The component was inside a
case which served as a protection from weather changes. Figure 2. Image Processing
III. RESULTS AND DISCUSSION
The face detection was firstly tested in this study. The
pedestrian detection was also tested and it was found that it
detected other objects other than the human body. The last was
the integration of the both detection.
The tables in this paper used the classification method
wherein the performances of the detections were tested. True

978-1-5090-0360-0/15/$31.00 2015 IEEE


8th IEEE International Conference Humanoid, Nanotechnology, Information Technology
Communication and Control, Environment and Management (HNICEM)
The Institute of Electrical and Electronics Engineers Inc. (IEEE) Philippine Section
9-12 December 2015 Water Front Hotel, Cebu, Philippines
positive and false negative were used to test for presence of the the null hypothesis. The difference computed in the pedestrian
target while true negative and false positive were used to test was high that resulted to high chi- square result and the
for the absence of the target. information could be insufficient. The true negative chi-square
accepted the null hypothesis which meant that true negative
Table 1. Summary of Data Gathered chi-square falls under the null hypothesis.
Face and Table 4. Chi-square of True Positive
Face Pedestrian
Pedestrian
True Positive 94 68 100 Face and
Face Pedestrian
False Negative 6 32 0 Pedestrian
True Negative 95 96 100 (O-E)2 36 1024 0
False Positive 5 4 0 2
Table 2. Values of True Positive (O-E) /E 0.36 10.24 0

Face and Table 5. Chi-square of True Negative


Face Pedestrian
Pedestrian
Observed 94 68 100 Face and
Face Pedestrian
Expected 100 100 100 Pedestrian
O-E -6 -32 0 (O-E)2 25 16 0
Table 3. Values of True Negative (O-E)2/E 0.25 0.16 0
Face and
Face Pedestrian IV. CONCLUSION
Pedestrian
Observed 94 68 100 The objectives of this study were successfully achieved.
Expected 100 100 100 The integration of the face and pedestrian detection was
O-E -6 -32 0 successful and implemented the system to the traffic light
system to change the signal transition of the traffic light. In
The null hypothesis was that the expected and observed integration of the two detections, there were two polygons
values are equal.This was compared later on when computing present in detecting a pedestrian. Based on the computed value
for chi-square. Computing for the overall rates of true positive, using the Chi-square, it concluded that the experiment or the
false negative, true negative and false positive, the following system can detect a pedestrians body and face and the success
were the computed percentage of each using the equation. rate of 87.33%. It also concluded that there were several factors

to considered that affects to detect a pedestrian, one of it was
TPR = = 0.8733 (1) the lighting of the environment and second was the

paraphernalia that a pedestrian was wearing such as umbrellas,
The success rate of the experiment was 0.8733. TPR caps, eyeglasses and also the color of the clothes that a
or true positive rate was the rate of correctly detecting the pedestrian wearing, etc.
pedestrian and their face.
V. RECOMMENDATION
Rate = (Total Column)/300 (2) The use of Histograms of Oriented Gradients or HoG
The rate of true positive was 0.8733. In false negative, the algorithm for pedestrian detection is recommended because
rate was 0.1267. The true negative rate was 0.97 and the false the HoG algorithm is much more precise to detect human
positive rate was 0.03. In the test of presence, the true positive figure because it gives higher performance to detect a human
rate showed a high percentage in detecting the targets. In the figure and the algorithm is based from the local intensity of
test of absence, the true negative rate was also high. the gradients of the image and also improving and research for
more up-to-date algorithm for face and pedestrian detection is
Table 2 showed the true positive of all detections. The recommended for more efficient and high performance of
difference was used to compute for the chi-square. The detection. The range of the camera used in gathering the data
difference in table 3 was also computed for the same purpose. is a short range distance to capture a human figure so by
Equation (2) was the chi square. designing the study with a high definition camera that has a
(O-E)2/E (3) long rage of distance and have the capability to have a night
vision to capture a human figure. Lastly, the future researchers
The total chi-square in table 4 was 10.5 while in table 5 the can formulate for more precise and high performance
total chi-square was 0.41. The true positive chi-square was algorithm that can perform face and pedestrian detection at the
10.5 and the true negative chi-square was 0.41. same time.
The significance level to be used for analysis was 0.75. ACKNOWLEDGMENT
The degree of freedom was 2 so the chi-square is 0.575. The
results showed that the true positive chi-square failed to accept The authors would like to acknowledge their advisers,
Engr. Jesus M. Martinez, Jr. and Prof. Cyrel O. Manlises in

978-1-5090-0360-0/15/$31.00 2015 IEEE


8th IEEE International Conference Humanoid, Nanotechnology, Information Technology
Communication and Control, Environment and Management (HNICEM)
The Institute of Electrical and Electronics Engineers Inc. (IEEE) Philippine Section
9-12 December 2015 Water Front Hotel, Cebu, Philippines
helping and guiding them. The authors would like
acknowledged the authors family and friends who helped and
support us as we finished this study.
REFERENCES
[1] Joel Schesser; Donald N. Tanner; , Hampton Roads
Traffic Surveillance and Control System IEEE
Transactions on Vehicular Technology, Vol. VT-29,
No.2, May 1980
[2] Yohan Dupuis; Xavier Savatier; Jean-Yves Ertaud;
PascualVasseur; , A New Approach for Face Detection
with Omnidirectional Sensors Proceedings of the
2011IEEE International Conference on Robotics and
Biometrics, December 2011
[3] Chandana S; , Real Time Video Surveillance System
Using Motion Detection Dept of Electronics and
Communication Engineering, DayanandaSagar College of
Engineering Bangalore, India
[4] Tom de Castella; , Does pressing the pedestrian crossing
button actually do anything? BBC News Magazine
September, 2013
[5] NavneetDalal; Bill Triggs, Histograms of Oriented
Gradients for Human Detection, Computer Vision and
Pattern Recognition, 2005. CVPR 2005. IEEE Computer
Society Conference on (Volume: 1), San Diego, CA,
USA, June 25, 2005
[6] Sheng Yang; Jiefa Wu; Lingling Zhang, A Fast
Pedestrian Detection Method Based on Simplified HoG
Descriptor, International Journal of Digital Content
Technology and its Applications (JDCTA), Volume 6;
Number 4, March 2012
[7] Ms. Jaya M. Jadhav; Ms. Deipali V. Gore, Introducing
Celebrities in an Images Using HAAR Cascaded
Algorithm, International Journal of Advanced Research
in Computer Science and Software Engineering (Volume
3, Issue 8), Pune India, August 2013

978-1-5090-0360-0/15/$31.00 2015 IEEE

Вам также может понравиться