Download Automatic License Plate Recognition Progress-Computer Sciences Applications-Project Report and more Study Guides, Projects, Research Applications of Computer Sciences in PDF only on Docsity!
i
Table of Contents
Table of Contents ................................................................................................................. i List of Figures ..................................................................................................................... ii
ii
- 1 Previous Work Abstract iii
- 1.1 PHASE 1 (LICENSE PLATE EXTRACTION)
- 1.1.1 Assumptions.......................................................................................................
- 1.1.2 Block Diagram
- 1.1.3 Preprocessing
- 1.1.4 Image Enhancement
- 1.1.5 Connected Component
- 1.1.6 Filtering Connected Components
- 1.1.7 Candidate Verification
- 1.2 SYSTEM CONFIGURATION
- 1.3 RESULTS
- 2 Current Work
- 2.1 Phase 2 (Character Recognition)
- 2.1.1 Character Segmentation
- 2.1.2 Character Recognition
- 2.2 RESULT
- 3 Conclusions
- 4 Future work
- Figure 1 Block diagram for Extraction Phase List of Figures
- Figure 2 Edge Detected Images
- Figure 3 Candidates after applying aspect ratio.
- Figure 4 Original Colored Images
- Figure 5 Extracted License Plates
- Figure 6 After applying constraints
- Figure 7 Extracted Characters
- Figure 8 Some Characters Templates
1 Previous Work
1.1 Phase 1 (License Plate Extraction)
In order to identify a vehicle by reading its license plate successfully, it is obviously necessary to locate the plate in the scene image provided by some acquisition system (e.g. in our case still camera). Locating the region of interest in the scene aids in dramatically reducing both the computational expense and algorithm complexity. For example, an image of resolution 640x480 which contains the region of interest (in this case a license plate) may account only 10% of the image area for region of interest. The technique used for extraction of plate is Edge Statistics and Morphology, based on the characteristics and shape of edges (Morphology).
1.1.1 Assumptions
To extract license plate from the image the assumptions are A visible boundary should be present between license plate and vehicle. License plate is of rectangular shape. Captured image contains only one and complete license plate attached to vehicle. License plate is clearly visible in the captured image.
1.1.2 Block Diagram
The block diagram for the extraction of license plate is shown below.
Figure 1 Block diagram for Extraction Phase
1.1.3 Preprocessing
In preprocessing the colored image is converted to grayscale. After that canny edge detector is used to detect the shape of license plate which is usually rectangular. Canny is used here because of its strength to detect all intensity changes which may result in some irrelevant details which can be removed in the filtering process to extract candidates discussed below.
Figure 2 Edge Detected Images
1.1.4 Image Enhancement
The edges detected by the Canny edge detector may not be connected in some of the cases, so it will be difficult to completely extract the License Plate if it is not fully edge detected. To over come this difficulty a thickening algorithm has been employed here. It thickens objects by adding pixels to the exterior of objects until doing so would result in previously unconnected objects being 8-connected. The result of applying this algorithm is that small gaps between edges are filled and we have 8- connected components in last.
Next candidates are further filtered by considering their length. We are considering images of vehicles in which license plate is attached to the vehicle than the length of the license plate must be less than the size of the image, using this constraint on the above filtered image using aspect ratio further reduces the number of candidates for the license plate.
1.1.7 Candidate Verification
The license plate of vehicles is a textured area as it contains characters on it. We can further verify it for smoothness. We have observed that from the selected candidates for license plate the license plate is an area which is least smooth as compared to the other area inside other objects such as wind screen, indicators etc. As a result we have a complete license plate in the end of this phase.
Figure 4 Original Colored Images
Figure 5 Extracted License Plates
1.2 System Configuration 800 Mhz Intel Original Processor. 256 MB RAM. Matlab 7.1.
1.3 Results
The system is tested on database of 100 images which include Pakistani and British License Plates. License Plate Recognition Rate is 91 %. Processing time is 2 seconds. 6% of plates were not correctly recognized (plates were half extracted). 3% of plates were not found.
Fig 5 Extracted License Plate.
Figure 6 After applying constraints
Figure 7 Extracted Characters
Character Recognition
Character recognition is divided in to two steps:
- Normalization of individual characters.
- Template matching.
In the first step the individual characters are normalized to a size of 40x40 using bilinear interpolation. In bilinear interpolation four nearest neighbor of a point are used for gray level assignment. If (x', y') denotes the coordinates of a point then v (x', y') is the gray level assigns to it using bilinear interpolation.
v (x', y') = ax'^ + by' + cx'y'^ + d Where the four coefficients are determined from the four equations in four unknowns that can be written using the four nearest neighbor of point (x',y').
Figure 8 Some Characters Templates
In the second step these characters are matched one by one with the template of characters stored in a database, and the template with the closest match is selected. The templates in database are made from the available vehicle images which are also normalized to a size of 30x30. This character database consists of 26 alphabets and 10 numeric templates. For template matching the cross correlation technique is used and the resultant value is in the range -1 to 1.The maximum value 1 indicates the closest or perfect match. The cross-correlation (or sometimes "cross-covariance") is a measure of similarity of two signals. It is a function of the relative time between the signals, is sometimes called the sliding dot product The principal use of correlation is for matching. In matching, f (x, y) is an image containing objects or regions. If we want to determine whether f contains a particular object or region in which we are interested, we let h (x, y) be that object or region (template). Then if there is a match then the correlation of two functions will be maximum at the location where h finds a correspondence in f. For discrete functions fi and gi the cross correlation is defined as
(f * g) (x) =ζ f *^ (t) g(x + t) dt
2.2 Result Character Recognition rate is 51.98 %. 278 characters were correctly recognized out of 535 characters. Frequently confused characters are (B, D, and O) (Z, 7) (P, R) (A, 4)
References
- Bai Hongliang and Liu Changping” A hybrid License Plate Extraction Method Based on Edge Statistics and Morphology” Proceedings of the 17th International Conference on Pattern Recognition , Volume: 2, ppt 831- 834 Vol.
- Hans A. Hegt, Ron J. De la Haye and Nadeem A. Khan” A High Performance License Plate Recognition System” in IEEE International Conference on Systems, Man, and Cybernetics, 1998.
- Shyang-Lih Chang, Li-Shien Chen, Yun-Chung Chung, and Sei-Wan Chen ,” Automatic License Plate Recognition ” in IEEE Transaction on Intelligent Transportation Systems, Vol5, 2004.
- John C.M.Lee, W.K.Wong and H.S.Fong, “Automatic Character Recognition for Moving and Stationary Vehicles and Containers” in International Joint Conference on Neural Networks, Volume: 4, ppt 2824-2828 ,1999.
- Tran Duc Duan, Tran Le Hong Du, Tran Vinh Phuoc and Nguyen Viet Hoang “ Building an Automatic Vehicle License-Plate Recognition System ” in Intl. Conf. in Computer Science, Can Tho, Vietnam, 2005.
- Rafel C. Gonzalez, Richard E. Woods, Steven L. Eddins, “Digital Image Processing Using Matlab” , pp 90-97, 373-377, 398-405,407- 418,478,483 2nd Edition.
- www.worldlicenseplates.com
