Conference PaperPDF Available

Accuracy of automatic number plate recognition (ANPR) and real world UK number plate problems

Authors:

Abstract

This paper considers real world UK number plates and relates these to ANPR. It considers aspects of the relevant legislation and standards when applying them to real world number plates. The varied manufacturing techniques and varied specifications of component parts are also noted. The varied fixing methodologies and fixing locations are discussed as well as the impact on image capture.
Accuracy of Automatic Number Plate Recognition
(ANPR) and Real World UK Number Plate Problems
Mike Rhead, Robert Gurney, Soodamani Ramalingam
School of Engineering and Technology
University of Hertfordshire
College Lane Campus
Hatfield, Herts AL10 9AB, UK
Email: s.ramalingam@herts.ac.uk
Neil Cohen
Home Office
Centre for Applied Science and Technology
Sandridge
St. Albans, Herts AL4 9HQ, UK
Abstract This paper considers real world UK number plates
and relates these to ANPR. It considers aspects of the relevant
legislation and standards when applying them to real world
number plates. The varied manufacturing techniques and varied
specifications of component parts are also noted. The varied
fixing methodologies and fixing locations are discussed as well as
the impact on image capture.
Key wordsANPR, number plates, optical character
recognition, National ACPO ANPR Standards (NAAS), ANPR
British Standards, Registration Marks Legislation, misreads,
missed reads.
I. INTRODUCTION
Automatic Number Plate Recognition (ANPR) within the
UK is a powerful tool and its policing purpose is to deny
criminals the use of our roads [1]. It is a major tool that is used
to improve the security and safety of the general public. In
standard use, ANPR is remarkably accurate. However, the
purpose of this work is to look at the occasions when ANPR
fails to capture the vehicle registration mark (VRM) correctly.
By reviewing the mode of failure it is hoped to be able to
improve the performance of ANPR. Some main causes of
ANPR failures area as follows:
Fixing Screws: Depending upon the character of the VRM
and the juxtaposition of a fixing screw misreads can occur.
Such misreads can be consistent or inconsistent and the paper
gives examples of this. An experiment whereby small diameter
anti-tamper screws were used to secure the number plate to the
vehicle is discussed as well as the relative improvement in
correct read rate.
Fixing Adhesives: Depending upon the method of
manufacture of the number plate, fixing adhesives can also
affect the retro-reflective layer and the photographic properties
of the number plate. This is caused by absorption of the
adhesive into the retro-reflective layer and subsequent impact
on its reflective properties.
Visible and near-infrared Spectrum: ANPR systems can
use both the visible and near infra-red spectrum. The
appearance of real world number plates can be significantly
different when comparing monochromatic or black and white
images from the visible spectrum and the near infra-red
spectrum.
The main cause of a misread read has been found to relate
to screws and screw caps where the fixing is within the
lettering of the VRM. The main failure mode for ANPR
cameras has been found to relate to fixings. The implications
for license plate manufacture [2], fixing methodology,
legislation and standards are discussed in the following
sections.
II. LITERATURE REVIEW
The authors of the paper work with Police ANPR systems.
Their work on other aspects of ANPR is well progressed and
further publications are planned. Countermeasures, cloned
plates and vehicles moving without number plates are an
inevitable aspect of this work and progress has already been
made in this area. However, the outcome of this aspect of
work will not be published nor put into the public domain.
This is a pioneering work; a literature review [3-6]
conducted by the authors revealed that there is no other
published real-world research. These papers assume ideal
experimental conditions. The papers that were identified
assumed ideal conditions.
Our initial work began in 2006 through field trials
determining the accuracy of hot lists. As a result of this work
changes were made to a number of hot lists to improve their
accuracy. In 2009, our attention turned to the cause of
misreads (not read correctly by ANPR systems) and by 2010
we were looking at the cause of missed reads (data missed
completely). During 2009 we also began researching the
NAAS Standard in depth. In 2010 we reviewed the British
Standard test for number plates, the details of the NAAS
standard and also the differences in performance of the
different ANPR cameras available and captured a lot of data
relating to these aspects. We are continuing to experiment with
camera settings (including software) and capture rate.
Field tests have been carried out for a range of ANPR
cameras under identical conditions to assess performance of
the camera, adjustment of camera setting and the performance
of the software within the camera [7]. Here, we are
considering fixings within a VRM. The results of this wide
ranging work are expected to help improve the accuracy of
ANPR to an even higher level.
III. FIXINGS
The following two images in Fig.1 are of the same number
plate. Fig. 1(a) was captured in the visible spectrum and Fig.
1(b) captured in the near infrared spectrum. The impact of the
screw cap can be seen between the two images. The screw cap
is much easier to see in the near infrared image.
Other unpublished work indicated that screws and screw
caps close to or within the lettering of the vehicle registration
mark could cause difficulty for optical character recognition
which in turn can lead to misreads or missed captures. The
impact of the screw fixing can be much greater under near
infrared imagery. This aspect is discussed in detail in the paper.
Figure 1 gives an example of the difference.
A. Initial Tests
Some of our initial (unpublished) work into misreads
indicated that ANPR misreads were caused by a number of
factors that could be exhibited by “real world” number plates.
The misreads were assessed and broken down into 5 categories
[Table I]. Our initial work looked at capture rates and misread
rates for vehicles giving a match to hot lists such as no
insurance. As an extension of this work the causes of 700
misreads to be studied in depth.
This initial work on 700 misreads suggested that number
plate fixings (screws and screw caps) accounted for 72.6% of
misreads. Results of analysis indicate that the top 10 characters
that accounted for 55.6% of incorrect reads and these are
related to screws or screw caps [Table II]. These results are
typical of the issue.
(A) VISIBLE SPECTRUM IMAGE (B) NEAR INFRARED IMAGE
FIGURE 1: DIFFERENCE BETWEEN VISIBLE SPECTRUM AND
INFRARED IMAGE CAPTURES
TABLE I: ATTRIBUTABLE CAUSE OF MISREAD
Percentage
Misreads
Cumulative
percent
Screw cap
72.6
72.6
Marks
23.0
95.6
Obscured
1.8
97.4
Broken
1.8
99.2
Illegal font
0.9
100.0
TABLE II: ATTRIBUTABLE CAUSE OF MISREAD
Character
Cumulate %
G
7.7
S
14.8
O
21.9
C
27.8
4
33.1
D
37.9
K
42.6
K
47.3
7
51.5
M
55.6
The fixing problem is more prominent at certain positions
on the Number Plate (NP). Fig. 2 indicates the position or
location of the misread character from this work. The post
2001 UK Number Plate regulation introduced a space between
the 4th and 5th characters. For the same 700 NPs, the character
at position 2 is the most commonly misread location (34%) and
position 5 is the second highest misread location (16%). These
two positions account for 50% of misreads caused by screws
and screw caps.
B. Further Field Tests
A number of vehicles in the UK have number plates that
begin with the letters OU [Figure 1]. Ten specific vehicles were
selected from a Police fleet each one having a screw/screw cap
fitted within the centre of the character U. Each had their
movements monitored over a two month period. There were
997 captures of which 713 (71.5%) were correctly read
indicating a misread rate of 28.5% for these plates. Of the 284
misreads the U was interpreted incorrectly as other characters
[Table III].
Five of the sample vehicles then had their fixing screw
replaced with smaller diameter anti-tamper screws in order to
try to reduce the footprint of the fixing. There were 359
captures for the 5 vehicles. Of these, 358 (99.7%) were
correctly read indicating a misread rate of 0.3%. The difference
in screw sizes are indicated in Figure 3.
From this test we infer that a simple change to the diameter
of the screw head used can significantly impact the accuracy of
the ANPR systems. The diameter of the standard fixing was
about 12mm and that of the anti-tamper fixing was 6mm. The
footprint area of the fixings went from 113mm2 to 28mm2. In
this test, our result indicated a 98.9% improvement in
accuracy!
FIGURE 2: POSITION OF MISREAD LETTER
TABLE III: STATISTICS OF INCORRECT READS
FIGURE 3: DIFFERENCE SCREW SIZES UNDER INFRA-RED
FIGURE 4: SCREW POSITIONS AFFECTING ANPR
INTERPRETATIONS
TABLE IV: RESULTS OF THE 1,238 VEHICLE SAMPLE
Screw fixings
Number in
category
Visible screws
966
Visible screws within the characters of
VRM
774
Outside the characters of VRM
192
No visible fixings
272
Table V Examples of misrepresented number plates
Further research identified screw positioning as being a key
factor causing mis-reads and resulting in incorrect
interpretations by ANPR systems as illustrated in Figure [3].
A) VISIBLE SPECTRUM B) NEAR INFRARED SPECTRUM
FIGURE 5. DIFFERENCE IN VISIBLE SPECTRUM VS IR
IMAGE CAPTURE
C. Confidence Measures
Having determined that there is a significant problem that
arises from screw fixings, further trials were undertaken to try
to determine the extent of the problem within the UK. To this
end, the fixing methodology was analysed for a sample of
1,238 vehicles, the results of which are as shown in Table IV.
D. Fixing Screws and Resulting Images in Visible and Near-
Infrared Spectra
The legislation requires that where the fixing screw goes
through a white background layer a white screw/screw cap is
used. Similarly a yellow fixing is used for the rear and a black
fixing goes through the actual lettering. The purpose of this is
to ensure that the number plate is easily read by the human eye
for identification and prosecution purposes.
The infrared spectrum used in ANPR applications to
capture number plates is just outside human vision [8,9]. To
give an indication of the difference two sets of number plates
captured in the visible spectrum are shown in [Fig. 5A] and the
same plates captured in the near infra-red spectrum are shown
in [Fig. 5B]. The fixing screws at the left and right hand edge
of the number plates appear as black dots in Figure 5B which
can contribute to mis-reads.
The reason for this is that the white plastic caps used as part
of the fixing are infrared absorbers and have no retro-reflective
properties. It would be illegal to use retro-reflective fixings on
a number plate. The difficulty faced as a result of this is that the
fixings within the VRM lettering appear as large black full
spots within the lettering and this can confuse the OCR
software depending upon the relative size and position of the
fixing with respect the character within the VRM.
E. Screw Caps and Spacing
Illegal use of screw caps and spacing in the visible
spectrum contravenes the statutory instrument. Table VI gives
a few examples of the illegal use of screws (as well as illegal
spacing). Such plates are illegal because they “deceive” the
eye as well as breaching number plate spacing regulations.
Such number plates do cause ANPR misreads.
Further work is also planned where we will look at the
impact of variable spacing using the standard UK number plate
Charles Wright font.
F. Fixings under Infrared
For the data considered in Section III B (Further Field
Tests) the reported misread rate of 28.5% was due to a screw in
the letter U and the size of the screw fixing [Fig. 7]. A further
analysis of 1,287 random number plates indicated that a large
number of plates have screws within the characters of the
VRM, typically 63% ± 4%. This does not imply that 63% of
number plates can be misread because of screw fixings; the
proportion of number plates giving misreads will be well below
this. Thus, one can easily conclude that the elimination of
screw fixings from number plates will improve the accuracy of
ANPR.
FIGURE 7: APPEARANCE OF A WHITE SCREW CAP FIXING ON A
WHITE RETRO-REFLECTIVE NUMBER PLATE UNDER NEAR
INFRARED
G. ANPR Infrared Images
The majority of number plate images used for optical
character recognition (OCR) are monochromatic images. The
images are normally captured by standard optical cameras
fitted with a visible spectrum and/or near infrared filters.
The infrared image is based upon the contrast between the
retro-reflective layer and characters of the VRM. Screws and
screw caps attenuate infrared and appear as black marks shown
in Figure 4 and this in turn can confuse OCR.
In summary, having considered the issue of one can see that
screw caps fitted within the type face can have a detrimental
effect on the ability of the human eye to recognise the VRM
correctly. At the same time, the situation can be even worse
under infrared. The legislation in this area requires clarification
as there are a number of ambiguous interpretations.
IV. INTERPRETATION OF THE STATUTORY INSTRUMENT AND
BRITISH STANDARD
A. Visible versus the near infrared spectrum
Greater clarity is required in the definition of the Act and
the Standard. The Act appears to cover infrared images because
it refers to “camera and film or any other device” and “true
photographic image”. There is no definition or case law with
respect to this and opinion is divided. The British Standard can
also be seen as ambiguous:
According to the British Standard, the luminous intensity in
a given direction of a source that emits monochromatic
radiation of frequency 555nm is defined by candela
abbreviated as cd.
The units of retro-reflectivity measurement of radiance is
given by
ρ = cd/lx/m2 (1)
where
cd candela, luminous intensity
lx lux, luminous flux per unit area
sr steradian, a unit of solid angle
m2 area in square meters
lux is defined by:
lx = 1 cd.strm2 (2)
Substituting (2) in (1), the retro-reflectivity unit of measure
becomes:
ρ = cd/cd.sr/m2/m2
The equation cancels and becomes
ρ = 1/sr
= per steradian
= per unit of solid angle
Thus, ρ is, in fact, a dimensionless unit and cancels to
become a per steradian.
The measurement relates to reflectance and is independent
of the wavelength of the radiation used for the test. That is, the
ratio of the total amount of radiation reflected by a surface to
the total amount of radiation incident on the surface.
The use of candela is taken by some to infer that the visible
spectrum is the required wavelength range because the unit
candela is mentioned. However, the British Standard [10] refers
to CIE 15 [11] and CIE 54 [12]. To measure the retro-
reflectivity of the characters and backing layer of a UK number
plate the standard requires illuminant A. I Illuminant A is based
on a tungsten filament bulb operating at a defined temperature
of 2,856 oK. Iluminant A has a defined wavelength range
(300nm to 830nm) and spectral power. The wavelength range
covered enters the near infrared spectrum. It is not clear if the
British Standard is required to be carried out at 555nm or over
the spectrum and spectral power defined in CIE 15 (300 to
830nm).
The key issue here relates to definition and interpretation of
Statutory Instrument No.561 [13]. The screw fixings clearly
interfere with the optical character recognition software. The
key question relating to these fixings is are they legal?
The requirements of the CIE standards were originally
applied to UK roads signs and markings with respect to driver
safety. They were subsequently incorporated into BS AU 145d
for UK number plates.
Some infrared ANPR systems operate at up to 940nm. As a
consequence the authors would like to see the defined range of
300-830nm amended within the British Standard. A range of
300-1,000nm would be preferred with reflectance measured as
an integral and at 5nm intervals. This would indicate if the
number plate had uniform reflectance properties over the
defined wavelength range.
B. Performance standards for fixings
The National Association of Chief Police Officers
Automatic Number Plate Recognition Standards (NAAS) [14]
notes minimum performance standards.
As has been demonstrated fixings can cause an ANPR
misread. If fixings do cause a misread they can be regarded as
an illegitimate UK plate. Under such circumstances these plates
can be discounted from the compliance test. This is not
unreasonable if the optimum performance possible (ideal
world) of the system is being considered. We would like to see
an additional measure whereby misreads attributable to fixings
are also declared. We propose that changes to UK number plate
legislation would tackle the main cause of ANPR misreads.
C. Variety in UK and Schengen number plates
Beyond the consideration of screw fixings there is a wide
variability in font, spacing and other marks in the registration
area of number plates within the UK. UK includes Northern
Ireland, Isle of Man and Channel Islands.
Under EU rules countries are, in effect, borderless for
member states. This is discussed more fully in another paper by
the authors [5]. There is an increased variety of number plates
being presented to ANPR systems and ANPR systems should
develop further to take this into account.
1) UK VARIETY
National ACPO ANPR Standard (NAAS) [6] notes that for
the avoidance of doubt, number plates from Northern Ireland,
Isle of Man and Channel Islands are regarded as UK number
plates.
The proportion of Northern Ireland, Isle of Man and
Channel Island number plates is comparatively small when
compared to standard UK plates. It is interesting to explore the
performance requirement for these plates under the NAAS
standard and the possible implications for ANPR
manufacturers and suppliers.
The Crown Dependencies of the Channel Islands and the
Isle of Man are outside the United Kingdom and European
Union, and have registration marks that differ from those used
in the UK.
Examples of the type and style of number plates included
are given in Fig. 6.
A) NORTHERN IRELAND B) JERSEY
C) GUERNSEY D) ISLE OF MAN
FIGURE 6: VARIETY IN NUMBER PLATES
2) ANPR algorithms to handle Schengen Community
number plates.
The series of images [App. A-B] relate to a sample of
Schengen community number plates and are given to illustrate
the challenge given to OCR readers in ANPR systems. The
difference in font, syntax, symbols and spacing pose a problem
to OCR utilised in the UK.
V. FUTURE WORK
ANPR captures are extremely accurate. This work has
determined that some number plates can be problematic for
ANPR cameras and that the main reason for a misread for a
UK number plate is caused by screw fixings. Screw fixings
within a VRM can confuse an ANPR system and the result
could be a complete failure to capture or a misread.
Future work we will consider includes:
Character spacing
Camera set up
Infrared illumination
Instantaneous traffic flow rates compared to
manufacturers hourly plate rate capacity
Comparative testing of different systems using a
defined data set
Various counter measure testing
Number plate manufacturing techniques
ACKNOWLEDGEMENTS
Dr. Vivienne Lyons - Home Office; Detective
Superintendent Paul Ealham - Hertfordshire Constabulary; Lisa
Gilmore - Department for Transport; Alastair Thomas - Home
Office; Mark Jones - National Policing Improvement Agency;
Frank Whiteley - former Chief Constable Hertfordshire.
References
[1] ACPO ANPR Steering Group (2005). Denying Criminals the
Use of the Roads. Report by the British Association of Chief
Police Officer’s ANPR Steering Group, March 2005.
[2] R. Gurney, M. Rhead, Private Discussions with British Number
Plate Manufaturer’s Association (BNMA), 2011.
[3] Z. Musoromy, S. Ramalingam. N. Beekoy, Edge Detection
comparison for License Plate Detection, Proc., 11th
International Conference on Control, Automation, Robotics and
Vision(ICARV2010), IEEE sponsored and organized by NTU,
Singapore, 7-10 Dec. 2010, pp.1133-1138.
[4] Z. Jeffrey, S. Ramalingam, High Definition License Plate
Detection Algorithm, Proc., IEEE SoutheastCon “Innovating for
a Better Tomorrow”, Mar.15-18, 2012, Orlando, Florida, US.
[5] Barry Watson and Karen Walsh, The Road Safety Implications
of Automatic Number Plate Recognition, Report, The Centre for
Accident Research & Road Safety, Queensland, Feb. 2008.
[6] Tran Duc Duan, Tran Le Hong Du, Tran Vinh Phuoc, Nguyen
Viet Hoang, Building an Automatic Vehicle License Plate
Recognition System, International conference in Computer
Science, Can Tho, Vietnam, Feb.21-24, 2005.
[7] Robert Gurney, Mike Rhead, Soodamani Ramalingam, Neil
Cohen, Working Towards an International ANPR Standard An
Initial Standard into the UK Standard, The 46th Annual IEEE
International Carnahan Conference on. Security Technology. 15-
18 October 2012, Boston, Massachusetts, USA, accepted.
[8] Some paradoxes, errors, and resolutions concerning the spectral
optimization of human vision BH Soffer & DK Lynch
http://www.phys.ufl.edu/~hagen/phz4710/readings/AJPSofferLy
nch.pdf, accessed 26 Mar 2012.
[9] H. Roy, “Wavelength considerations”, Instituts für Umform-
und Hochleistungs,
http://web.archive.org/web/20071028072110/http://info.tuwien.a
c.at/iflt/safety/section1/1_1_1.htm, accessed 26 Mar.2012.
[10] BS AU 145d:1998, British Standards Institution, 15 January
1998, ISBN 0-580-28985-0,
[11] The International Commission of Illumination (CIE)
Publication No. 15 (CIE), 1986 Colorimetry
[12] The International Commission of Illumination (CIE)
Publication No. 54 (CIE) Retro-reflection definition and
measurement
[13] “Statutory Instrument No. 561 The Road Vehicles (Display of
Registration Marks) Regulations,” 2001,
[14] The National ACPO ANPR Standards (NAAS) National
Policing Improvement Agency June 2012
APPENDIX A: LIST OF SCHENGEN STATES
APPENDIX B: SOME EXAMPLES OF SCHENGEN
COMMUNITY NUMBER PLATES
France
Spain
Portugal
Croatia
Hungary
Czech Republic
Belgium
Lithuania
Poland
Romania
Bulgaria
Latvia
Germany
Austria
Estonia
Austria
Belgium
Bulgaria
Cyprus
Czech Republic
Denmark
Estonia
Finland
France
Germany
Greece
Hungary
Iceland
Italy
Latvia
Lithuania
Luxembourg
Malta
Netherlands
Norway
Poland
Portugal
Romania
Slovak Republic
Slovenia
Spain
Sweden
... An automatic number plate recognition for real-world application for the United Kingdom number plate recognition in [19]. Several real-world aspects are incorporated in the system including the screw fixing, font variability, spacing, color, and different symbols present on the number plates. ...
Article
Full-text available
Non-standard license plates are a part of current traffic trends in Pakistan. Private number plates should be recognized and, monitored for several purposes including security as well as a well-developed traffic system. There is a challenging task for the authorities to recognize and trace the locations for the certain number plate vehicle. In a developing country like Pakistan, it is tough to have higher constraints on the efficiency of any license plate identification and recognition algorithm. Character recognition efficiency should be a route map for the achievement of the desired results within the specified constraints. The main goal of this study is to devise a robust detection and recognition mechanism for non-standard, transitional vehicle license plates generally found in developing countries. Improvement in the character recognition efficiency of drawn and printed plates in different styles and fonts using single using multiple state-of-the-art technologies including machine-learning (ML) models. For the mentioned study, 53-layer deep convolutional neural network (CNN) architecture based on the latest variant of object detection algorithm-You Only Look Once (YOLOv3) is employed. The proposed approach can learn the rich feature representations from the data of diversified license plates. The input image is first pre-processed for quality improvement, followed by dividing it into suitable-sized grid cells to find the correct location of the license plate. For training the CNN, license plate characters are segmented. Lastly, the results are post-processed and the accuracy of the proposed model is determined through standard benchmarks. The proposed method is successfully tested on a large image dataset consisting of eight different types of license plates from different provinces in Pakistan. The proposed system is expected to play an important role in implementing vehicle tracking, payment for parking fees, detection of vehicle over-speed limits, reducing road accidents, and identification of unauthorized vehicles. The outcome shows that the proposed approach achieves a plate detection accuracy of 97.82% and the character recognition accuracy of 96%.
... Just in case of written document pictures, presence of skew, touching or overlapping text line(s) makes this method a true challenge to the research worker. The SVM classifier achieves 89.35% text lines with success from the aforesaid databases severally[9]. ...
Article
Full-text available
The development of the recognition techniques has been led to an easy way to get useful information from digital images; this information can be used in printed word recognition from digital documents. This attempts to develop a powerful computing system for scanning the text from pictures. Currently, there is a large demand for storing data found in paper documentand so later reusing this data. In this paper, document images are used to find out the printed word within. These images are contained phrases and long texts, they obtained from Internet. The proposed method applies two preprocessing stages for preparing these document images. Then, images segmentation is used by scanning the image contents for looking the contents of the documents line by line and word by word. The challenges concerned during this the font characteristics of the characters in paper documents and quality of pictures, which overcame by unifying these limitations. The use of SURF descriptor was successful to describe the shape of each word by a number of interest points as word features; each interest point is a vector of 128 elements. These interest points are enrolled to gathers for different words in database file. The SVD classifier decision is based on SURF features, in which the average word recognition score was 99.514%, in which the amount of the varying this score is about ±0.036, which indicates the acceptable recognition results achieved by the SURF based word recognition method.
... However, as discussed earlier, a big problem in ALPR is the accuracy of detection [12]. The same thing faced in the UK and real-life issues in detecting ALPR studied were found [13]. The main reason behind this is the failure during the extraction of the characters in the proper way. ...
Article
Full-text available
Automatic license plate recognition (ALPR) is a widely used system in developed countries to track and classify vehicles. Contrariwise, least developed countries such as Bangladesh are having issues like rotated, slightly faded license plates. Additionally, there are multi-line license plates where usually two rows are used to interpret vehicle numbers that create complexity during character extractions. These also inflate problems for the sequential sorting of the characters after extraction, let alone the rotated one. Thus, no well-established ALPR system is discoverable for those countries. Character recognition techniques might be available for some popular languages; still, they are not becoming effective due to the lack of pre-processing of the images. Therefore an effort has been made to process images of the rotated multi-line license plates to retrieve their original orientations and then extract and order all the characters sequentially accordingly. This algorithm is expected to help develop the ALPR system to distinguish individual characters even from the inferior quality license plates. Consequently, any standard recognition technique might be applied to unique character images obtained from this algorithm to recognize characters developed for any particular language accordingly.
... However, as discussed earlier, a big problem in ALPR is the accuracy of detection [12]. The same thing faced in the UK and real-life issues in detecting ALPR studied were found [13]. The main reason behind this is the failure during the extraction of the characters in the proper way. ...
Article
Full-text available
Automatic license plate recognition (ALPR) is a widely used system in developed countries to track and classify vehicles. Contrariwise, least developed countries such as Bangladesh are having issues like rotated, slightly faded license plates. Additionally, there are multi-line license plates where usually two rows are used to interpret vehicle numbers that create complexity during character extractions. These also inflate problems for the sequential sorting of the characters after extraction, let alone the rotated one. Thus, no well-established ALPR system is discoverable for those countries. Character recognition techniques might be available for some popular languages; still, they are not becoming effective due to the lack of pre-processing of the images. Therefore an effort has been made to process images of the rotated multi-line license plates to retrieve their original orientations and then extract and order all the characters sequentially accordingly. This algorithm is expected to help develop the ALPR system to distinguish individual characters even from the inferior quality license plates. Consequently, any standard recognition technique might be applied to unique character images obtained from this algorithm to recognize characters developed for any particular language accordingly.
... digunakan, warna yang digunakan pada plat nomor kendaraan, penggunaan gambar pada latar belakang dan plat nomor kendaraan yang memiliki standar resmi sendiri membuat pengenalan plat nomor kendaraan cukup mengalami tantangan yang sulit [8]. Pada penelitian ini menggunakan metode jaringan syaraf tiruan, karena metode ini mampu menunjukkan efisitasnya sebesar 98,17% [9] dan mampu lebih cepat dibandingkan dengan metode lainnya [10]. ...
Article
In certain applications, native surface patterns can be used in place of speckle patterns in digital image correlation (DIC). This paper explores the feasibility of using text as a native speckle pattern in DIC. Five text speckle patterns are tested in three different scenarios: a rigid body translation test, a rigid body rotation test, and an out of plane bending test. The patterns are benchmarked against a sixth, random speckle pattern applied using traditional DIC speckling methods. Rigid body translation tests are additionally performed on text patterns with varying font types and line spacings. In general, text patterns have good contrast, but low density as line spacing increases. Measurement uncertainty for the text patterns was comparable to measurement uncertainty in the random speckle pattern. Results from these tests show that while text patterns cannot be expected to perform better than a traditional DIC speckle pattern, text patterns can be effective speckle patterns in situations where already present on a specimen and applying a traditional speckle pattern is difficult.
Article
Full-text available
In today's world there has been a huge increase in the use of digital technology and various methods of capturing images are available. These images may contain important textual data that the user may need to digitally edit or archive. This can be done by optical character recognition using the Tesseract OCR engine. The main important concept behind this technology is something called OCR-Optical Character Recognition. Using OCR, we can search and recognize text in electronic documents and easily convert it to readable text. It converts the text of electronic documents into a relative ASCII character and, if the document is handwritten, OCR uses the database to recognize which character it is and resolve it with the utmost precision. In this article we have discussed and analyzed different methods for recognizing text from images. The purpose of this review article is to summarize the known methods for a better understanding of the reader.
Article
This paper describes two studies that aimed to explore the impacts of pedestrianisation or road closures on traffic displacement, travel behaviour and the phenomenon of ‘disappearing traffic’. The first study surveyed residents whose travel routes were affected by a small-scale localised pedestrianisation scheme in the centre of a town. The second measured the traffic impacts of a temporary closure of a strategic bridge in a city centre. In the first case, the pedestrianisation produced no change in the modal shares of travel of residents. Drivers continued to drive to the same locations by longer routes. In the second case, the closure caused some traffic displacement and increased journey times but also reduced traffic volumes in both the immediate area and across the city. This paper concludes by discussing the remaining knowledge gaps on disappearing traffic, made more pressing by the decisions of authorities to reallocate road space during the coronavirus disease 2019 crisis.
Article
Full-text available
There is a significant increase in the utilization of digital technology in the current world, and various methods are available for people to capture images. Such images may contain necessary textual data that the user may need to edit or store digitally. This whole process is done using Tesseract which is a part of Optimal Character recognition (OCR)The essential central concept behind this technology is something called OCR-Optical Character Recognition. With the OCR's help, we can search and recognize the text in electronic documents and quickly convert them into human-readable text. It transforms electronic documents' text into related ASCII characters. If the form is a handwritten one, then the OCR uses a database to recognize its character and try to solve it to its highest accuracy. In this paper, we have reviewed and analysed different methods for text recognition from images.
Conference Paper
Full-text available
Due to a huge number of vehicles, modern cities need to establish effectively automatic systems for traffic management and scheduling. One of the most useful systems is the Vehicle License-Plate (VLP) Recognition System which captures images of vehicles and read these plates’ registration numbers automatically. In this paper, we present an automatic VLP Recognition System, ISeeCarRecognizer, to read Vietnamese VLPs’ registration numbers at traffic tolls. Our system consists of three main modules: VLP detection, plate number segmentation, and plate number recognition. In VLP detection module, we propose an efficient boundary line-based method combining the Hough transform and Contour algorithm. This method optimizes speed and accuracy in processing images taken from various positions. Then, we use horizontal and vertical projection to separate plate numbers in VLP segmentation module. Finally, each plate number will be recognized by OCR module implemented by Hidden Markov Model. The system was evaluated in two empirical image sets and has proved its effectiveness (see section IV) which is applicable in real traffic toll systems. The system can also be applied to some other types of VLPs with minor changes.
Article
Full-text available
This paper describes a background noise elimination technique introduced for License Plate (LP) detection algorithms specifically designed for High Definition (HD) images to deal with the surplus data they contain. The images are firstly enhanced using a robust method, followed by the application of morphological operators and histogram percentile autonomous thresholding for removing background noises keeping the resulting image grey. Finally, greyscale edge detection based segmentation is applied to extract the candidate regions of interest. Experiments on thousands of images show an improvement not only in LP detection in HD images, but also in edges processing time, which compensates for the additional time due to background noise elimination. The proposed algorithm is also tested on Standard Definition (SD) images where higher LP detection success is observed on SD images with complex background scenes.
Conference Paper
Full-text available
The detection of license plate region is the most important part of a vehicle's license plate recognition process followed by plate segmentation and optical character recognition. Edge detection is commonly used in license plate detection as a preprocessing technique. This paper compares the performance of the image enhancement filters when used in edge detection algorithms combined with connected component analysis to extract license plate region. The experimental comparison of Canny, Kirsch, Rothwell, Sobel, Laplace and SUSAN edge detectors on gray scale images shows that Canny yields high plate detection of 98.2% tested on 45,032 UK images containing license plates at 720×288 resolution captured under various illumination conditions. The average processing time of one image is 56.4 ms.
Article
Full-text available
This submission has been prepared in response to the Parliamentary Travelsafe Committee’s Inquiry into Automatic Number Plate Recognition Technology. This submission will outline how automatic number plate recognition (ANPR) works, its history and application, as well as some key privacy and data management issues. After outlining evaluations that have been undertaken of ANPR and highlighting some limitations, the submission will examine the implications of the technology for road safety. The primary focus of this submission is on the potential of ANPR to both detect and deter a range of illegal road user behaviours including unlicensed driving, the driving of unregistered and uninsured vehicles, speeding, and the noncompliance of heavy vehicle drivers with driving hour regulations and novice drivers with provisional licence requirements. To this end, the submission will review different options for deploying ANPR, the risks associated with these different approaches, and how the likely road safety benefits of the technology could be maximised. To assist in this process, a set of principles is presented that can be used to assess the potential road safety benefits of ANPR technology.
Conference Paper
This paper examines the use of the UK National AC PO ANPR Standard (NAAS) as the "de facto" technical standard applied in many international countries. It considers the requirement for a standard and examines the effectiveness of the current NAAS and questions its fitness for purpose. The need for accuracy is discussed in terms of both tackling terrorism, serious crime and other law enforcement investigations alongside the need to protect citizens from unwarranted infringement of their privacy as a result of ANPR misreads. The causes of inaccurate ANPR read data are examined in more detail and recommendations made as to how improvements could be introduced to minimise the risk of misreads and "missed" reads. This paper recommends future parameters of measurement and provides examples of gaps between the current standards and existing legislation. Laboratory and field testing was carried out to gain a better understanding of the factors that affect the performance of ANPR systems. These tests were carried out under a variety of weather and lighting conditions. The results of this work have led to further testing to better understand the optimum conditions for number plate capture by a variety of ANPR systems. Additional testing has been carried out using "hard to read" number plates with a number of differing characteristics such as illegally spaced characters, illegal fonts, screw caps that interfere with infrared imaging and defects in the construction of the number plate itself (whether created inadvertently at the point of manufacture or subsequently caused by damage /wear and tear / weather conditions). The first author is a UK police officer and, like his senior analyst colleague who is the second author, has wide experience in testing and developing ANPR systems. The authors have been commissioned by the UK Home Office to carry out post graduate ANPR research at the University of Hertfordshire.
Denying Criminals the Use of the Roads. Report by the British Association of Chief Police Officer's ANPR Steering Group
  • Acpo
  • Steering
ACPO ANPR Steering Group (2005). Denying Criminals the Use of the Roads. Report by the British Association of Chief Police Officer's ANPR Steering Group, March 2005.
Private Discussions with British Number Plate Manufaturer's Association (BNMA)
  • R Gurney
  • M Rhead
R. Gurney, M. Rhead, Private Discussions with British Number Plate Manufaturer's Association (BNMA), 2011.
High Definition License Plate Detection Algorithm Innovating for a Better Tomorrow
  • Z Jeffrey
  • S Ramalingam
Z. Jeffrey, S. Ramalingam, High Definition License Plate Detection Algorithm, Proc., IEEE SoutheastCon " Innovating for a Better Tomorrow ", Mar.15-18, 2012, Orlando, Florida, US.
Working Towards an International ANPR Standard – An Initial Standard into the UK Standard, The 46th Annual IEEE International Carnahan Conference on
  • Robert Gurney
  • Mike Rhead
  • Soodamani Ramalingam
  • Neil Cohen
Robert Gurney, Mike Rhead, Soodamani Ramalingam, Neil Cohen, Working Towards an International ANPR Standard – An Initial Standard into the UK Standard, The 46th Annual IEEE International Carnahan Conference on. Security Technology. 15-18 October 2012, Boston, Massachusetts, USA, accepted.
The Road Safety Implications of Automatic Number Plate Recognition, Report, The Centre for Accident Research & Road Safety
  • Barry Watson
  • Karen Walsh