چكيده
Abstract
In recent years Intelligent Transportation Systems (ITS) have become an
important research area and this is because of importance of their applications in
real world problems. ITS is the application that incorporates electronic, computer,
and communication technologies into vehicles and roadways for monitoring
traffic conditions, reducing congestion, enhancing mobility, and so on. One of the
problems in ITS, is recognizing of car models. Vehicle type recognition, if solved
accurately, is beneficial for authentications checking, police camera control
systems on crossings to match the number-plate against the car make and tracking
the special car. The proposed methods will provide valuable situational
information for law enforcement units in a variety of civil infrastructures. Various
researches have been done on license plate recognition (LPR) systems but make
and model recognition (MMR) is an unexplored problem. By unification of LPR
and MMR systems valuable and useful information can be mined.
Proposed algorithm begins with detecting cars in the image, and then regions of
interests are localized for feature extraction. After the extraction of valuable
features that are different in various models of vehicle, vehicle’s model in the
input image is determined using a classifier. Image dataset used for training and
test includes 290 images from side and back view of various vehicle models in
RGB color map. Experimental results show that the features which are used for
car model recognition in our method are highly effective. Ensemble of classifiers
in which each classifier classifies cars based on different features is also used.
Results also confirm our prediction that ensemble of classifiers is more powerful
in car model recognition. the average recognition rate of our proposed system is
96.2% for images in our test set which contain most of car models in Iran.
Key Words
Vehicle Detection, License Plate Localization, Vehicle Make and Model
Recognition, Computer Vision, Intelligent Transportation System
