CHARACTERS AND NUMBERS RECOGNITION FROM LICENCE PLATES OF REPUBLIC OF SERBIA USING NEURAL NETWORKS
Dušan Stevanović · Knowledge International Journal · 2017
In this paper it has been described the Optical Character Recognition (in further text OCR) method, which is widely used in today’s world. The work is based on a model of using the OCR method, recognizes the letters and numbers from the image that user gave. The network is trained with the dataset of letters and numbers using the built-in functionality in Matlab, known as Neural Network Toolbox. The image that user assigns is analyzed and its content is compared with the training set, and then the final results are displayed. Recognized letters and numbers are used for various purposes. In today's world, which is at a high level of technological development, there is a tendency to get different demands, needs or solutions of the problems - quickly and easily. An increasing number of works and texts are published in digital format, so access to essential information is very fast. On the other hand, it is simply impractical to copy all texts, files, documents in order to get their digital duplicate and edit them.The process of transforming physical documents into digital was a real challenge, but also great relief for digital technology users. However, soon there was a need for changing the content of documents stored in the digitalization process in a format that does not allow such a type of operation (formats such as .jpg, .gif, .png do not allow the possibility of any change in content). Different approaches are proposed for identifying letters and numbers so that, if necessary, the entire text could be easily updated or edited. Significant successes were recorded in this intention. One of the most common methods used for this purpose is Optical Character Recognition (OCR) in combination with networks.The ways of using the OCR method are different, and they can vary from simple transfer of material into digital form, their changes, through recognizing someone's handwriting, to recognizing texts from pictures and photos. In this paper we will describe the problem of recognizing the content from the registered license plates of Republic of Serbia. The practical usage of the OCR method is very often realized by training and using networks.Neural represent a relatively new concept in analyzing and processing data. By the term neural networks in this paper we will refer to artificial networks.Neural are widely used, both in technical and social sciences. The more precise the prediction of the values of the relevant variables is, the more successful training of network is. In addition to the use of for characters recognition, they are also very popular and practical in the field of classification and clustering. The performance of in performing these three functions defines them as an extremely powerful tool.