30-08-2014, 10:41 AM
Vision-Based Automated Parking System
Vision-Based.pdf (Size: 1 MB / Downloads: 49)
Abstract
- This paper describes an approach to overcome a
situation of monitoring and managing a parking area using a
vision based automated parking system. With the rapid
increase of cars the need to find available parking space in
the most efficient manner, to avoid traffic congestion in a
parking area, is becoming a necessity in car park
management. Current car park management is dependent on
either human personnel keeping track of the available car
park spaces or a sensor based system that monitors the
availability of each car park space or the overall number of
available car park spaces. In both situations, the information
available was only the total number of car park spaces
available and not the actual location available. In addition,
the installation and maintenance cost of a sensor based
system is dependent on the number of sensors used in a car
park. This paper shows a vision based system that is able to
detect and indicate the available parking spaces in a car park.
The methods utilized to detect available car park spaces were
based on coordinates to indicate the regions of interest and a
car classifier. This paper shows that the initial work done
here has an accuracy that ranges from 90% to 100% for a 4
space car park. The work done indicated that the application
of a vision based car park management system would be able
to detect and indicate the available car park spaces
1. INTRODUCTION
Parking of cars in a parking area is becoming a
difficult task as the number of cars increases while the
number of parking spaces is finite. As a result, people
would spend a certain amount of time looking for
parking space and thus cause a situation where the
traffic would be slowed down and cause congestion.
The situation of looking for parking space and traffic
congestion in parking areas is due to the fact that the
information of available parking spaces is not readily
available to the people looking for parking spaces. As
such different approaches have been used to develop a
car park management system such as wireless sensor
network system [1] and a vision system [2].
2. BACKGROUND
In recent years, parking a car has become a serious
problem in large cities with increasing rate of private
vehicles [3]. With the emerging problem of parking
cars, the ordinary parking system that does not provide
any information about available parking areas would
not be able to handle the problem effectively. The
typical car park system would only be able to provide
information of available parking locations or another
system would require human resources to determine
and provide information about the location of the
available parking locations. These types of parking
systems would only provide minimal information on
the available parking locations and would not be able
to handle the parking issues effectively. As such these
systems would get the drivers to search the parking
areas on their own and thus create a problem where
there would be too many cars in the car park area.
In order to address the problem of parking
effectively, sensors can be utilized to detect and
provide information on the location of available
parking areas. Among the implementation of sensor
based parking system is a wireless sensor system [1].
This system would utilize sensors in each parking
space would provide information on the status of each
car park locations but the cost of installing sensors in
each parking bay might prove to be prohibitive [4] as
the cost of installing sensors would increase with the
increase in the number of parking bays or area
A. Feature Extraction
In this paper, Haar-like features were used in the
detection of features detected in input videos to
determine the presence of a car within a parking bay.
Haar-like features use the changes in contrast values
between adjacent groups of pixels rather than actual
pixel values to determine common Haar features within
an image [5]. The primary purpose of using Haar-like
features would be for easier classification rather than
raw pixel values as the Haar-like features are done in
windows of 24x24 pixels [6]. The Haar-like features
are typically used to determine the information of a
region rather than the raw pixel values. From the
utilization of a group of Haar-like features, the
determination of features and computed values are
used as input into a decision tree classifier for
identi fication.
B. Object Identification
The basis of object identification in this paper was
based on the utilization of Haar-like features. The tool
for object classification was developed in an open
source library called Open Computer Vision Library
(OpenCV) [5]. In order to train the object classification
algorithm, two sets of images are required to train the
classifier. One set of input images would be images
that contain the object to be detected, which can be
called as positive images, and another set of images
that do not contain the object, that are called negative
images. In the training of the classifier, the location of
the object within the positive image including the
height and width of the object need to be specified.
3. RESULTS AND DISCUSSION
In the development of a car park management
system to detect available parking space or bays, one
factor needs to be taken into account which is the cost
of implementation. Systems that determine the
availability of car park spaces require sensors to
determine the status of the parking spaces would incur
cost in obtaining sensors, preparing and maintaining
the infrastructure of the parking system [1,4]. The
work described in this paper was a vision based car
park management system that utilizes a web camera.
The car park system developed was based on object
classification and coordinate method to determine the
status of car parking spaces.
The object classification for the car park system
used was haartraining that was available in OpenCV.
The object classifier is used to detect the presence of a
target object within an input image. The object
classifier is trained with both positive and negative
images. The positive images used in training the
classifier were images with cars from all types and
angles. Samples of positive input images can be seen in
Figure 4. The negative images used in the classifier
training were images that did not contain any cars in
the image.
4. CONCLUSION
The results obtained from the testing of the vision
based car park system on model cars and car parks, it
indicates that the probability of the utilization of a
camera or vision based system to monitor car parks is
feasible. This is seen in the results that indicate the
accuracy in determining the presence of a car in a
parking bay or the information that can be provided on
the location of the available parking bays. The
development of the vision based car park system shows
the feasibility of utilizing simple cameras such as web
cameras to monitor car parks. However, the utilization
of such cameras has indicated some weaknesses in
terms of the accuracy of the detection due to the
limitations of the camera.
In order to overcome some of the weakness that
was observed in the development of this vision based
system, improvements can be made in the type of
camera used where the web cameras can be tied in
together with CCTV cameras. In addition, the
coordinate method used in this paper in selecting the
specific parking locations has limited the usage by
making the each parking bay location fixed and thus
limits the camera to be at a fixed location. As such, the
usage of character recognition can be used to
determine exact parking bay that is available and thus
frees up the camera to be able to change its position
and be able to pan and tilt to have a larger view of the
car park area and thus use less cameras. In order to
have greater ability in detecting the presence of a car
within the region of interest, the object or car classifier
would have to be trained with a larger set of positive
and negative images.