27-08-2016, 10:18 AM
1445766280-AutomaticLicensePlateRecognition.PDF (Size: 486.04 KB / Downloads: 5)
Abstract. A License plate recognition (LPR) system can be divided into the following
steps: preprocessing, plate region extraction, plate region thresholding,
character segmentation, character recognition and post-processing. For step 2, a
combination of color and shape information of plate is used and a satisfactory
extraction result is achieved. For step 3, first channel is selected, then threshold
is computed and finally the region is thresholded. For step 4, the character is
segmented along vertical, horizontal direction and some tentative optimizations
are applied. For step 5, minimum Euclidean distance based template matching
is used. And for those confusing characters such as '8' & 'B' and '0' & 'D', a special
processing is necessary. And for the final step, validity is checked by machine
and manual. The experiment performed by program based on aforementioned
algorithms indicates that our LPR system based on color image processing
is quite quick and accurate.
Introduction
The automatic identification of vehicles has been in considerable demand especially
with the sharp increase in the vehicle related crimes and traffic jams. It can also play a
crucial role in security zone access control, automatic toll road collection and intelligent
traffic management system. Since the plate can identify a car uniquely, it is of
great interest in recent decade in using computer vision technology to recognize a car
and several results have been achieved [2-14].
A typical LPR system can be divided into the following modules: preprocessing
(including image enhancement and restoration), plate region extraction, plate region
thresholding, character segmentation, character recognition and post-processing (validity
checking). The first two modules, which only concern the shape and back/fore
ground color of a plate and irrespective of character set in a plate, are the front end of
the system. Module 4 and 5, on the contrary, are related to character set in a plate and
regardless of the shape and back/fore ground color of a plate, so they are the back end
of the system. Module 3, however, should take the shape and back/fore ground color
of a plate as well as character set in a plate into consideration. Therefore, it is hard to
say which end it can be categorized into.
To develop an automatic recognition system of a car plate, a stable recognition of a
plate region is of vital importance. Techniques such as edge extraction [1][6], Hough transformation [7] and morphological operations [8] have been applied. An edgebased
approach is normally simple and fast. However, it is too sensitive to the unwanted
edges, which may happen to appear in the front of a car. Therefore, this
method cannot be used independently. Using HT is very sensitive to deformation of a
plate boundary and needs much memory. Though using gray value shows better performance,
it still has difficulties recognizing a car image if the image has many similar
parts of gray values to a plate region, such as a radiator region [11][12]. Morphology
has been known to be strong to noise signals, but it is rarely used in real time
systems because of its slow operation. So in recent years, color image processing
technology [4][5] is employed to overcome these disadvantages. First, all of the plate
region candidates are found by histogram. After that, each one is verified by comparing
its WHR (Width to Height Ratio), foreground and background color with current
plate standard and eliminated if it is definitely not of plate region. And finally, for
each survivor, an attempt to read plate information is made by invoking the back end.
In the back end, first channel is selected and the plate region is thresholded in the
selected channel. And then, each character is extracted by histogram and some optimizations
such as the merge of unconnected character (i.e. Chuan, or ), the removal
of space mark, frame and pin, the correction of top and bottom coordinates in y
direction and tilt correction are done during this phase. Next, each character is recognized
by using minimum Euclidean distance based template matching since it's more
noise tolerant than structural analysis based method [2][3]. And for those confusing
characters, '8' & 'B' and '0' & 'D', for instance, a special processing is necessary to
improve the accuracy. Finally, validity checking is performed against vehicle related
crimes.