License plate recognition (LPR) technology has great importance for the development of Intelligent Transportation Systems by automatically identifying the vehicles using image processing and pattern recognition techniques. Conventional LPR systems consist of license plate detection (LPD), character segmentation (CS) and character recognition (CR) steps. Successful detection of license plate and character locations have vital role for proper LPR. Most LPD and CS techniques in the literature assume fixed distance and orientation from the vehicle to the imaging system. Hence, application areas of LPR systems using these techniques are limited to stationary platforms. However, installation of LPR systems on mobile platforms is required in many applications and algorithms that are invariant to distance, orientation, and illumination should be developed for this purpose. In this thesis work, a LPD algorithm that is based on multi-scale vertical edge density feature, and a character segmentation algorithm based on local thresholding and connected component analysis operations are proposed. Performance of the proposed algorithm is measured using ground truth positions of the license plate and characters. Algorithm parameters are optimized using recall and precision curves. Proposed techniques for each step give satisfying results for different license plate datasets and algorithm complexity is proper for real-time implementation if optimized.

Background: Every country has seen an increase in motorcycle accidents over the years due to social and economic differences as well as regional variations in transportation circumstances. One common mode of transportation for those in the middle class is a motorbike.  Every motorbike rider is legally required to wear a helmet when driving a bike. However, some people on bikes used to ignore their safety, which resulted in them violating traffic rules by driving the bike without a helmet. The policeman tried to address this issue manually, but it was ineffective and proved to be quite challenging in practical circumstances. Therefore, automating this procedure is essential if we are to effectively enforce road safety. As a result, an automated system was created employing a variety of techniques, including Convolutional Neural Networks (CNN), the Haar Cascade Classifier, the You Only Look Once (YOLO), the Single Shot multi-box Detector (SSD), etc. In this study, YOLOv3 and Haar Cascade Classifier are used to compare motorcycle helmet and license plate detection.  Objectives: This thesis aims to compare the machine learning algorithms that detect motorcycles’ license plates and helmets. Here, the Haar Cascade Classifier and YOLO algorithms are used on the US License Plates and Helmet Detection datasets to train the models. The accuracy is obtained in detecting the helmets and license plates of the motorcycles and analyzed.  Methods: The experiment method is chosen to answer the research question. An experiment is performed to find the accuracy of the models in detecting the helmets and license plates of motorcycles. The datasets utilized for this are from Kaggle, which included 764 pictures of two distinct classes, i.e., with and without a helmet, along with 447 unique license plate images. Before training the model, preprocessing techniques are performed on US License Plates and Helmet Detection datasets. Now the datasets are divided into test and train datasets where the test data set size is considered to be 20% and the train data set size is 80%. The models are trained using 80% pre-processed training datasets and using the Haar Cascade Classifier and YOLOv3 algorithms. An observation is made by giving the 20% test data to the trained models. Finally, the prediction results of these two models are recorded and the accuracy is measured by generating a confusion matrix.   Results: The efficient and best algorithm for detecting the helmets and license plates of motorcycles is identified from the experiment method. The YOLOv3 algorithm is considered more accurate in detecting motorcycles' helmets and license plates based on the results.  Conclusions: Models are trained using Haar Cascade and YOLOv3 algorithms on US License Plates and Helmet Detection training datasets. The accuracy of the models in detecting the helmets and license plates of motorcycles is checked by using the testing datasets. The model trained using the YOLOv3 algorithm has high accuracy; hence, the Neural network-based YOLOv3 technique is thought to be the best and most efficient.

Made available in DSpace on 2016-03-15T19:37:29Z (GMT). No. of bitstreams: 1 Luiz Angelo D Amore.pdf: 3689058 bytes, checksum: 8476274d8f5220a2a7978da28a4a4f3d (MD5) Previous issue date: 2010-08-13 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / The work presented here shows a robust method for license plate detection. The term robust in this work is directly related to the efficacy of the system as an automated locator of license plates without human intervention and considering specific characteristics of image acquisition and license plate features. The proposed method is based on the specify features of the digits found on the Brazilian license plates. Although the method was designed for the Brazilian license plate pattern it can be easily adjusted to other patterns. The results obtained using the proposed method showed a better performance than that of other academic approaches and even of commercial systems. / Os sistemas automáticos de reconhecimento de placas veiculares têm como principal função a identificação de veículos a partir de imagens digitais, com aplicações nas áreas de segurança pública e privada. Neste trabalho são apresentadas técnicas de processamento de imagens com o objetivo de desenvolver um método robusto para a segmentação de placas veiculares em condições complexas de aquisição. O termo robusto neste trabalho é relacionado diretamente à eficácia do sistema quanto à localização automática das placas veiculares sem intervenção humana, considerando características específicas das imagens e placas. O método proposto é baseado nas especificidades dos dígitos localizados nas placas brasileiras. Embora o método tenha sido projetado para o padrão de placas brasileiro, pode ser facilmente ajustado para outros padrões. Os resultados obtidos com o método proposto mostram um desempenho melhor que outras abordagens acadêmicas, ou mesmo de sistemas comerciais.

<p>Disertacija prikazuje novi algoritam za detekciju i izdvajanje registarskih tablica iz slike vozila koristeći fazi 2D Gaborov filtar. Parametri filtra: orijentacija i talasna dužina su fazifikovani u cilju optimizacije odziva Gaborovog filtra i postizanja dodatne selektivnosti filtra. Prethodno navedeni parametri dominiraju u rezultatu filtriranja. Bellova i trougaona funkcija pripadnosti pokazale su se kao najbolji izbor pri fazifikaciji parametara filtra. Algoritam je evaluiran nad vi&scaron;e baza slika i postignuti su zadovoljavajući rezultati. Komponente od interesa su efikasno izdvojene i postignuta značajna otpornost na &scaron;um i degradaciju na slici.</p> / <p>The thesis presents a new algorithm for detection and extraction of license plates from a vehicle image using a fuzzy two-dimensional Gabor filter. The filter parameters, orientation and wavelengths are fuzzified to optimize the Gabor filter&rsquo;s response and achieve a greater selectivity. It was concluded that Bell&rsquo;s function and triangular membership function are the most efficient methods for fuzzification. Algorithm was evaluated on several databases and has provided satisfactory results. The components of interest were efficiently extracted, and the procedure was found to be very noise-resistant.</p>

This thesis aims at proposing vehicle license plate detection and recognition algorithms, suitable for vehicle re-identification. Simple urban traffic analysis system is also proposed. Multiple stages of this system was developed and tested. Specifically - vehicle detection, license plate detection and recognition. Vehicle detection is based on background substraction method, which results in an average hit rate of ~92%. License plate detection is done by cascade classifiers and achieves an average hit rate of 81.92% and precision rate of 94.42%. License plate recognition based on Template matching results in an average precission rate of 60.55%. Therefore the new license plate recognition method based on license plate scanning using the sliding window principle and neural network recognition was introduced. Neural network achieves a precision rate of 64.47% for five input features. Low precision rate of neural network is caused by small amount of training sample for some specific license plate characters.