VEHICLE LICENSE PLATE DETECTION AND RECOGNITION
Table Of Contents
<p> TABLE OF CONTENTS </p><p>ACKNOWLEDGEMENTS ............................................................................................................... II </p><p>LIST OF FIGURES ...................................................................................................................... V ABSTRACT .................................................................................................................................... VII</p><p>
Chapter 1
INTRODUCTION AND BACKGROUND ................................................................... 1 </p><p> 1.1 Research Topic and Objectives ................................................................................................. 1 </p><p> 1.2 Challenges ......................................................................................................................... 2 </p><p> 1.3 Background and Related Work ............................................................................................... 3 </p><p> 1.3.1 License Plate Detection ....................................................................................................... 4 </p><p> 1.3.2 Character Segmentation ....................................................................................................... 13 </p><p> 1.3.3 License Plate Recognition ..................................................................................................... 15</p><p> 1.4 Overview of Our Methods ...................................................................................................... 16 </p><p>2 LICENSE PLATE DETECTION ................................................................................................ 18</p><p> 2.1 Scanning Window ............................................................................................................... 18 </p><p> 2.2 HOG Features .................................................................................................................... 20 </p><p> 2.2.1 Feature Extraction Procedure (for Dense-HOG) .............................................................. 21 </p><p> 2.2.2 Implementation .............................................................................................................. 24 </p><p> 2.3 Support Vector Machine ......................................................................................................... 24 </p><p>2.4 Non-maximum Suppression ..................................................................................................... 26 </p><p> 2.5 Refinement of the Algorithm................................................................................................. 29 2.5.1 Edge Information.............................................................................................................. 29</p><p> 2.5.2 Scale Adaption..................................................................................................................... 30</p><p> 2.6 Results and Discussion ........................................................................................................... 30</p><p> 2.7 Summary ............................................................................................................................. 31 </p><p>3 LICENSE PLATE RECOGNITION ............................................................................................ 32 </p><p> 3.1 License Plate Alignment Using Color Information................................................................... 34 </p><p> 3.1.1 License Plate Alignment Without Angles.............................................................................. 34 </p><p> 3.1.2 License Plate Alignment With Angles............................................................................ 38 </p><p> 3.2 Plate Binarization Using K-means Clustering........................................................................... 40 </p><p> 3.2.1 K-means Clustering.............................................................................................................. 41 </p><p> 3.3 Character Segmentation Using an Innovative Histogram-based Model ................................. 43 </p><p> 3.4 Digit Characters and Capital Letters Recognition Using Simple Robust Features .................. 48 </p><p> 3.4.1 Using Bag-of-words Model: Voting Schemes........................................................................ 49 </p><p> 3.4.2 Using SVM Classifier.......................................................................................................... 53 </p><p> 3.5 Data set and Results ............................................................................................................ 54 </p><p> 3.6 Summary.................................................................................................................................... 57</p><p> 4 REAL TIME EMBEDDED SYSTEM ............................................................................................ 58 </p><p> 4.1 Hardware Part...................................................................................................................... 58 </p><p> 4.2 Software Part.........................................................................................................................59 </p><p> 4.2.1 The Main Board Side............................................................................................................. 59 </p><p> 4.2.2 The Child Board Side........................................................................................................ 60 </p><p> 4.3 Implementation..................................................................................................................... 62 </p><p> 4.3.1 Kernel Module................................................................................................................... 62 </p><p> 4.3.2 Main Board Program.......................................................................................................... 63 </p><p> 4.3.3 Socket and TCP................................................................................................................. 63 </p><p> 4.3.4 Child Board Program........................................................................................................ 64 </p><p> 4.4 Results and Discussion.................................................................................................................................... 64 </p><p>5 CONCLUSIONS AND FUTURE WORK ................................................................................. 66 </p><p>REFERENCES .............................................................................................................................. 67 </p>Project Abstract
AbstractVehicle license plate detection and recognition have become essential tasks in various applications such as traffic management, law enforcement, and parking systems. This research focuses on the development of an efficient system for automatic license plate detection and recognition using deep learning techniques. The proposed system consists of two main stages license plate detection and character recognition. In the first stage, a deep learning model is employed to detect license plates from images captured by surveillance cameras or other sources. The model is trained on a large dataset of annotated images to accurately locate license plates under various lighting and environmental conditions. Once the license plate region is detected, a character recognition module is applied to extract the alphanumeric characters from the plate. This module utilizes convolutional neural networks (CNNs) to recognize individual characters with high accuracy. The CNN model is trained on a dataset of labeled characters to learn the patterns and variations in different fonts and styles. To improve the overall performance of the system, various pre-processing techniques are applied to the input images, such as image enhancement, noise reduction, and normalization. These techniques help improve the quality of the input data and enhance the robustness of the deep learning models. The system is designed to be scalable and adaptable to different types of vehicles and license plate formats. It can handle variations in plate size, orientation, and background clutter, making it suitable for real-world applications with diverse scenarios. Experimental results demonstrate the effectiveness of the proposed system in accurately detecting and recognizing license plates in challenging conditions. The system achieves high detection and recognition rates on a benchmark dataset, outperforming existing methods in terms of accuracy and efficiency. Overall, this research contributes to the advancement of automatic license plate detection and recognition systems by leveraging deep learning techniques and innovative approaches. The system shows promising results in real-world applications and can be further optimized and extended to address specific requirements in different domains.
Project Overview
1.0 INTRODUCTION AND BACKGROUND
1.1 Research Topic and Objectives
License Plate Recognition (LPR) is a problem aimed at identifying vehicles by detecting and recognizing its license plate. It has been broadly used in real life applications such as traffic monitoring systems which include unattended parking lots, automatic toll collection, and criminal pursuit [5]. The target of this thesis is to implement a vehicle retrieval system for a Chinese surveillance camera, by detecting and recognizing Chinese license plates. It will be useful for vehicle registration and identification, and therefore may further contribute to the possibility of vehicle tracking and vehicle activity analysis.
The proposed method includes two main steps:
1) License Plate Detection: Using SVM classifier with HOG features based on a sliding window scheme, scan possible regions detected by edge information, and obtain license plate candidates. Then apply Non-Maximum Suppression (NMS) to finalize the plate locations.
2) License Plate Recognition: The detected license plate will be aligned first, after which its pixels can be successfully clustered by k-means into two classes: background pixels , and the foreground pixels, e.g., the pixels of the characters. The plate is segmented afterwards, into character patches that will be recognized using SVM classifier individually.
1.2 Challenges
The first challenge is plate variation. The plate can be various in location, quantity, size, color, font, occlusion, inclination, and plates may contain frames or screws [1]. The second challenge is environmental variation which includes change in illumination and background. Weather conditions, lighting conditions, and even camera conditions may contribute to the difficulty of this problem. For Chinese license plates, the font is fixed except vanity plates designed by individuals, which is very rare. Fig.1.1 shows some Chinese license plates from several provinces. The challenge from the font is relieved while all the other challenges like variation in plate location, plate quantity, its size, color, occlusion, and inclination still remains. The weather condition where the images in the dataset are captured is various too. Images are captured in the daytime as well as at night; the weather can be sunny or it can be rainy.
1.3 Background and Related Work
License plate recognition (LPR) system contribute to applications such as traffic surveillance, traffic law enforcement, automatic toll collection, vehicle parking identification, and vehicle access control in a restricted area. Typically, an LPR system is composed of License plate detection (LPD), license plate character segmentation (LPS) and License plate character recognition (LPR). License plate detection is commonly the first procedure in a LPR system. It aims at locating the license plate, which provides the following LPR procedure with accurate region information. Instead of processing every pixel in the input image of the system, which is very time consuming, license plate detection is a necessary process before license plate recognition. Methods on LPD can be classified into such categories based on the color level: 1) Binary Image processing; 2) Gray-Level Processing; 3) Color Processing; 4) Classifiers [3]. While based on the information used to detect the license plate, the LPD methods can be categorized into six classes: 1) using edge information; 2) using global image information; 3) using texture features; 4) using color features; 5) using character features; 6) combing two or more features [1].
1.3.1 License Plate Detection
A. Using Edge or Boundary Information.
a. Filters
In [6]-[9], Sobel filter is used to detect edges, by which the boundaries of license plates are represented due to color transition between the license plate and the car body. Two horizontal lines are located when performing horizontal edge detection, and two vertical lines are located when performing vertical edge detection. The rectangle is fixed when two set of lines are located both at the same time. While in 2003, F. Kahraman [2] et al. applied Gabor filters to detect license plate regions which achieved a good performance when images are of a fixed angle. Gabor filters are helpful in analyzing textures as they are sensitive to textures with different scales and directions. b. Edge Detection Based on the intuitive that the license plate is of some shape, most likely rectangular, whose aspect ratio is known, methods are commonly used to extract plates from all possible rectangles. Edge Detection methods are such ones to find the rectangles [10]-[13]. Edges can be detected only vertically or horizontally, and can be statistically analyzed to determine license plate candidate regions [12]. A fast vertical edge detection algorithm (VEDA) was proposed in [14] for license plate detection, which is declared to be faster than Sobel operator by around seven to nine times
Blazingprojects Mobile App
📚 Over 50,000 Project Materials
📱 100% Offline: No internet needed
📝 Over 98 Departments
🔍 Software coding and Machine construction
🎓 Postgraduate/Undergraduate Research works
📥 Instant Whatsapp/Email Delivery
Related Research
Implementing Virtual Reality Technology for Enhanced Learning in Computer Science Ed...
The project topic "Implementing Virtual Reality Technology for Enhanced Learning in Computer Science Education" focuses on the integration of virtual ...
Implementing Virtual Reality Technology for Enhanced Learning in Computer Education...
The project topic, "Implementing Virtual Reality Technology for Enhanced Learning in Computer Education," focuses on exploring the integration of Virt...
Developing an Interactive Online Learning Platform for Computer Science Education...
The project "Developing an Interactive Online Learning Platform for Computer Science Education" aims to address the growing need for innovative and en...
The Impact of Gamification on Student Engagement and Learning in Computer Science Ed...
The project topic "The Impact of Gamification on Student Engagement and Learning in Computer Science Education" focuses on exploring the effects of in...
Design and Development of an Interactive Online Learning Platform for Computer Scien...
The project topic "Design and Development of an Interactive Online Learning Platform for Computer Science Education" focuses on the creation of a digi...
Implementing Virtual Reality Technology for Enhancing Computer Science Education...
The project topic "Implementing Virtual Reality Technology for Enhancing Computer Science Education" focuses on the utilization of virtual reality (VR...
Developing a Virtual Reality Tool for Enhancing Computer Science Education...
The project titled "Developing a Virtual Reality Tool for Enhancing Computer Science Education" aims to explore the potential of virtual reality (VR) ...
Implementing Virtual Reality Technology in Computer Science Education...
Virtual Reality (VR) technology has emerged as a powerful tool with significant potential to transform the field of education, including computer science educat...
Implementation of Augmented Reality in Computer Science Education...
The project topic "Implementation of Augmented Reality in Computer Science Education" focuses on the integration of augmented reality (AR) technology ...
