Car License Plates Recognition for Intelligent Radar System Noura A. Semary1,
Ahmed A. Shouaib2, Ahmed H. Ezzat3
Faculty of Computers and Information, Menofia University, Egypt
Faculty of Computers and Information, Menofia University, Egypt
1
[email protected]
Abstract— Automatic Number License Plate (ANLP) recognition basically depends on using optical character recognition (OCR) technology to read LPs on vehicles. Such system is essential for traffic control, theft control, electronic toll collection, parking lots access and/or billing. In this project, a proposed Automatic Arabic numbers license plate (AANLP) recognition system is applied to Egypt new Arabic LPs. The proposed system recognizes both Arabic numerals as well as Arabic alphabets. The system goes through many preprocessing steps in order to produce segmented characters of the LPs images. The feature extraction step uses the count of black pixels from the horizontal projection profiles in addition to the black pixel distributions in divided zones of the character image. The recognition is performed separately based on different regions therefore, applying the recognition process to specific regions of the LPs speeds up the processing due to the limited number of comparisons. The system has been tested on 70 images of both day and night images. The accuracy of the system reached %92.5 for day images and %83.3 for night images what is good enough as a proof of concept. The accuracy could be increased by addressing the system limitations.
2
[email protected] [email protected]
3
II. METHODS AND MATERIALS A typical ANLP automatic recognition system goes through several major stages as shown in Figure 2 and each stage may contain several steps [1]. The image acquisition can be done using a digital camera or a video camera with a frame grabber to select a frame. Following image capture, some preprocessing is required to prepare the image of the LP for the LP extraction stage. Next, the characters are detected then segmented and features are extracted from each character. Finally, the characters go through the recognition stage.
Index Terms—Car Plates, OCR, character recognition.
I. INTRODUCTION Automatic number license plate recognition was invented in 1976 and the first working prototype system was implemented in 1979 in the UK [1]. ANLP recognition is used in many applications such as access control, traffic violations, law enforcements, border control, electronic billing … etc., such systems are based on image processing and computer vision techniques [2]. Several recognition techniques have been applied to Automatic LPs recognition systems these are[2-8]: • Template matching. • Structural and statistical analysis. • Fuzzy logic. • NNs approaches. LPs found in Arab countries contain the name or abbreviation of the name of the country in Arabic. Our LP recognition system concerns on new LPs that is shown in Figure 1.
Fig. 1 Egypt Arabic License Plates
Fig. 2 Block diagram of ALP recognition system Proposed Method
Our proposed AANLP recognition system consists of basic image processing steps [9] that will be discussed in details in this section: A. RGB to Binary conversion: The captured image is converted to grayscale using the intensity equation; where the intensity I = (R+G+B)/3. Single threshold is then used to separate white objects in the image. Figure 3 presents at the left a LP image from our dataset captured at night . while presents in the right the binary image obtained by single threshold th = 145.
Erosion : The erosion of A by B , is written by AϴB which is defined as: AϴB= {ω: 𝐵𝜔 ⊆ A} (1) Dilation : The dilation of A by B , is written by A B which is defined as: A B= ⋃𝑏𝜖𝐵 𝐴𝑏 (2) Opening : The opening of A by B , is written by A○B which is defined as an erosion operation followed by a dilation operation: (3) Fig. 3 Image binarization. (left) Original (right)the binary image
Using a specific structure element allows objects with specific size to be eliminated (Figure 6).
B. Filling the image holes : As the plate numbers are the only black objects rely on a white background, the numbers could be extracted by subtracting the original image from a filled version. Holes filling operation is performed on the binary image to fill the LP like in Figure 4.
Fig. 6 Filter by opening operation
E. Rotation Estimation: According to the position of both the camera and the moving car, the LP could be rotated (tilted). By estimating the rotation angle, an inverse rotation function could be applied to the image to make all the objects on the base. Figure 7 illustrates the process. Fig. 4 Filling the image holes
C. To Remove Unused object (Difference) By getting the difference between filled image and the original binary image, the unwanted white objects will be eliminated and the numbers will be remained. (Figure 5)
Fig. 7 rotation estimation
The rotation operator performs a geometric transform which maps the position P1 = (x1,y1) of a pixel in an input image onto a position P2 = (x2,y2) in an output image by rotating it through a user-specified θ about an origin O =(x0,y0), (Figure 7). Such that: 𝜃 = tan−1
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗ ‖𝑃 1 𝑃2 ‖ ⃗⃗⃗⃗⃗⃗⃗⃗1 ‖ ‖𝑂𝑃
Fig. 5 Extracting characters
D. Morphological opening: To remove all white objects containing less than a specified number of pixels, a morphological opening operation with a small a structuring element is applied where:
Fig. 8 After correcting the rotation angle
(4)
F. Cropping : by knowing the text position in the image crop the text from the image to reduce unwanted objects in the image G. Object extraction (connected component) By tracing connected components, each character could be detected from the image and a rectangle can be drawn around each character to get each character separately (Figure 9)
Fig. 9 (left) After Cropping , (right) After connected objects extraction
H. Optical Character Recognition In our system, template matching strategy is used for OCR purpose. Once the characters could be split successfully (Figure 10), the characters are resized into fixed size as well as the templates in our dataset. The matching procedure is followed like stated in the flowchart in Figure 11:
Fig. 10 Characters after splitting
Template matching determines the similarities between template database and input text. After correlation step take the max value of comparison and take the index of this character in the template matrix and print it in the text file.
Fig. 11 Flow chart of Template matching
III. RESULTS AND DISCUSSION In this section, the results obtained by our system will be depicted. A. Dataset collection A dataset of 70 images (40 day images and 30 night images) has been captured using different digital cameras and under different conditions; illumination, angles, fog, colors and distances. Figure 12 presents some examples of our dataset images.
Fig. 12 different examples of our dataset images
B. Results Table 1 presents the system true and false rates. The system could recognize 92.5% of our dataset while it failed to recognize 3 images of 40 day images and 5 out of 30 night images due to noise (Figure 13.1st row, Figure 14.1st row), far distance (Figure 13.2nd row) or camera flash effect (Figure 14. 2nd and Figure 13. 3rd rows). Table. 1 System Results
Image type Day Images Night Images
True 37 25
False 3 5
Accuracy 92.5% 83.3%
captured in different daylight conditions. Preprocessing steps including thresholding and morphology operations are used for refining the numbers regions. Plate rotation could be estimated to inverse the rotation and return the characters to its original angle as possible. Template matching is used for performing OCR stage after preprocessing. System results showed the ability to use such system for intelligent radar and road control systems. C. Future Work There are many challenges in ALP recognition like as the uncontrolled effect of illumination, plate rotation, old or dirty plates. Also the system will be adapted to real time applications and video shots have to be processed instead of images[10]. REFERENCES [1] S. Chang, L. Chen, Y. Chung, and S. Chen, Automatic License Plate Recognition. IEEE Transactions on Intelligent Transportation Systems, Vol. 5, No. 1, March 2004. [2] V. Sharma, P. C. Mathpal, A. Kaushik, Automatic license plate recognition using optical character recognition and template matching on yellow color license plate. International Journal of Innovative Research in Science, Engineering and Technology, Vol. 3, Issue 5, May 2014. [3] R. Dhruw and Dh. Roy. "Automatic Number Plate Recognition System." International Journal of Computer Science and
Mobile Computing, Vol.3 Issue.7, July- 2014, pg. 6-12 Fig. 13 Day images with false results
Fig. 14 Night images of false results
CONCLUSION In this paper we have proposed simple car license plates recognition system dedicated for Egyptian car plates. The system could successfully recognize more than 20 test plates
(2014). [4] R. Singh, and N. Randhawa. "Automobile Number Plate Recognition And Extraction Using Optical Character Recognition." International Journal of Scientific & Technology Research Vol. 3, issue 10, October 2014. [5] H. Mokayed, et al. "Car Plate Detection Engine Based on Conventional Edge Detection Technique." The International Conference on Computer Graphics, Multimedia and Image Processing (CGMIP2014). The Society of Digital Information and Wireless Communication, 2014. [6] A. Badr, M. Abdelwahab, A. Thabet, and Ahmed Mohamed Abdelsadek. "Automatic number plate recognition system." Annals of the University of Craiova-Mathematics and Computer Science Series38, no. 1 (2011): 62-71. [7] H. Sarukhanyan, S. Alaverdyan, and G. Petrosyan. "Automatic Number Plate Recognition System." Proceedings of the 7th International Conference on Computer Science and Information Technologies. 2009, pp. 347-350 [8] S. Mohammad Alavi and M. Alipour Varmazabadi, “Car Plate Detection with Neural Network Based on Particle Swarm Optimization (PSO)” Advances in Environmental Biology, 9(3) February 2015, Pages: 774-779. [9] M. Petrou and C. Petrou, Image Processing: The Fundamentals. 2010 edition. John Wiley & Sons, Ltd. ISBN: 978-0-470-745861 [10] C. Pornpanomchai, and N. Anawatmongkon. "Thai License Plate Recognition System From a Video Stream." Silpakorn University Science and Technology Journal Vol. 8(2) (2014), pp 40-50.
