1 TERNOPIL ACADEMY OF NATIONAL ECONOMY INSTITUTE OF COMPUTER INFORMATION TECHNOLOGIES SMART VEHICLE SCREENING SYSTEM USING ARTIFICIAL INTELLIGENCE METHODS Presenters: Volodymyr Turchenko Vasyl Koval
The Structure of the License Plate Recognition System 2 License Plate Recognition System License Plate Extraction Subsystem Character Recognition Subsystem
The Approach for License Plate Recognition 3 Images Capture of of the the Car 1 2 3 Deblurring of of Captured Images Extracting of of the the License Plate Chars Extracting on on License Plate 4 5 Recognizing of of License Plate Characters First Subsystem Second Subsystem
4 Deblurring of Captured Images N M Image Image Template CF RF TF + = 1) 1)*( ( ) 1 ( ) ( 1 1 + = = = N M n m n m RF M m N n 1) ( 1)* ( 1) ( ) ( 1 1 + = = = N M n m n m CF M m N n + < + > =. )) 2( ) 1( *( 2 1 ; 2 1 ) 2( ; 2 1 ) 1( ) 2( else j i j i Treashold TF TF if j i Treashold TF TF if j i j i Third stage First stage Second stage
Image fusion result 5 Blurred images Original image Deblurring result
The Approach for License Plate Extraction 6 Image Thresholding Regions of of Possible Plates Selecting License Plate Identification Extracting of of the the License Plate
Image Tresholding Technique 7 3( i j) = 1 if 2( i j) > Treshold 0 else Deblurred image Thresholded image
Technique of Possible Plates Regions Selecting 8 Thresholded image Procedure of Region Numbering (i-1j) ( i j) = Number if (ij-1) (ij+1) (ij) (i+1j) ( i ( i 1 j) = 1 ( i + 1 j) = 1 j) = 1 ( i j 1) = 1 ( i j + 1) = 1 Image matrix with numbered objects
Technique of License Plate Identification 9 Parameters to to Region Region for for Identification: Total Total pixels pixels in in region region Region Region height height Region Region width width Existence of of license license characters in in the the region region Identified Region with License Plate
Chars Extracting on License Plate 10 Preprocessing of of License Plate Image Character Selecting Character Resizing Characters Extracting on on License Plate
Chars Extracting on License Plate 11 Preprocessing of license plate License plate cutting from the image Thresholding of license plate image Inverting of license plate image Filtering the license plate Character selecting Numbering (labeling) of license plate object Extracting the character taking into account the following criteria: Width of the selcted object High of the selected object Total sum of white pixels in the object Character resizing Cutting of the character if criteria are true Resizing a character region to the set values of width and high
Recognizing of License Plate 12 Neural Network Structure 11 1 2 2 21 1 2 ( IW p + b ) a = f ( LW a b ) 2 21 1 11 1 2 f ( LW f ( IW p + b ) b ) 1 1 a = f + 2 a = + Output Output calculation calculation dx = mc*dxprev + lr*mc*dperf/dx Training Training technique: technique: Gradient Gradient descent descent backpropagation backpropagation method method with with momentum momentum and and adaptive adaptive learning learning rate rate
Experimental Results 13 Training progress of neural network Character classification of neural network
Dependences Between Noise Level and Correct Classification of License Plate Characters 14
Conclusions 15 The purpose of this paper is to investigate the possibility of automatic recognition of vehicle license plate. We can improve quality of the vehicle image using fusion technique then extract the license plate and isolate characters contained on the plate and finally identify the characters on the license plate using artificial neural network. The experimental results have shown the ability of neural network to recognize correctly characters on license plate with probability of 95% in presence of noise with 50% density. The proposed approach of license plate recognition can be implemented by police to detect speed violators parking areas highways bridges or tunnels. Also the prototype of the system is going to be integrated and tested as part of the sensor network being developed by Ayers Island LLC as part of their Intelligent Island system.