1.0 Introduction During the recent years, image processing based vehicle license plate localisation and recognition has been widely used in numerous areas:- a) Entrance admission b) Speed control Modelling, Simulation and Computing Laboratory (msclab) c) Intelligent Transportation Systems (ITS) d) As a core for intelligent infrastructure (toll& parking payment) Automatic License plate recognition (ALPR) is a form of automatic vehicle identification. It is an image processing based technology used to identify vehicles by only their license plate. The ALPR system consists of three steps: a) Vehicle localisation b) License plate localisation c) Character segmentation and recognition 1
2.0 Objective Modelling, Simulation and Computing Laboratory (msclab) ALPR systems have been physically utilised in many facilities such as parking lots, security control of restricted areas, traffic surveillance and so on. Regardless of the fact that the license plate can be intentionally altered in fraud situations, license plates still remains the elementary vehicle identifier. The aim of this project is to develop a tracking and localisation system of moving vehicle license plate via Signature Analysis. 2
Modelling, Simulation and Computing Laboratory (msclab) 3.1 Motion Detection & Pre-processing Input fed to system Extract frame Process colour image greyscale=0.299*red+0.587*green+0.114*blue Background subtraction frame i frame i-4 > T h Binarised Median filtering (2x2) 3
Modelling, Simulation and Computing Laboratory (msclab) 3.1 Motion Detection & Pre-processing (cont) Vehicle Localisation An algorithm will search for both minimum and maximum pixels of the labeledcomponents horizontally and vertically (along x-axis and y-axis) 4
3.2 License Plate Localisation Modelling, Simulation and Computing Laboratory (msclab) Signature Analysis not limited to any specific colour or significant edges to locatealicenseplateinamixedimage. License plate text forms a dense region of vertical strokes, irrespective of standpoint change and inclination. This unique pattern of vertical strokes is identified as the signature of a license plate. The Signature can be represented by a perceptual characteristic of texture in terms of repetition of components in a vertical direction (peaks) and the gap between two peaks. The number of peaks generated is counted to determine whether it falls betweenthethresholds.inthisproject,thethresholdissetto6and15. 5
Modelling, Simulation and Computing Laboratory (msclab) 3.2 License Plate Localisation (cont) Adaptive Searching - used to increase the efficiency and save the computational time. Instead of re-searching for the signature in each iteration (re-processing segment by segment on each frame), the first successfully calculated vehicle s height will be used as the reference height in the next iteration. Thus, in the next iteration, reference height will be compared with the vehicle s current height. If value of the reference height is smaller than the current height, this implies that the vehicle is moving forward, and vice versa. 6
Modelling, Simulation and Computing Laboratory (msclab) 4.0 Experimental Results and Discussions License plates can come in many patterns such as one-row license plates, tworows license plates, framed or frameless license plates etc. The frame of a license plate can play an important role in locating and segmenting the region of interest in an image. The system designed using a signature analysis approach works well with onerow license plates and frameless license plates. However, the major problem in the proposed background subtraction failed to distinguish two vehicles moving simultaneously. 7
Modelling, Simulation and Computing Laboratory (msclab) 4.1 Limitation - Two Vehicles Moving Simultaneously 8
4.2 Successful Results Modelling, Simulation and Computing Laboratory (msclab) 9
5.0 Conclusion Modelling, Simulation and Computing Laboratory (msclab) There are many methods that can be used for dynamic tracking and localisation of a vehicle s license plate. In this paper, the flow of procedure is: Motion Detection & Pre-processing Image differencing & filtering Vehicle localisation License Plate Localisation Adaptive searching Signature Analysis Character Segmentation Isolating connected components It can be concluded that the multiple vehicles motion detection and license plate localisation via isotropic dilation is not one hundred percent precise. The system needs to be heightened in the future so that it may run smoothly to overcome the limitations 1 0