Identificação óptica de caracteres de placas de trânsito.

Abstract

This study aims to develop a system to identify an automobile license plate receiving as an input an image containing a picture of the vehicle. The system should detect the plate in the image and recognize the characters. In order to do this, a bibliographic research was carried out in computer vision, digital image processing and pattern recognition. The Visual Studio programming environment, the library OpenCV and the software Tesseract were the computational tools used to accomplish the objective. For the detection of the plate in the image, a Haar training was applied. After detection, the image of the plate was separated from the rest and processed and then submited to the character recognition phase. The process of detection reached a precision of 50% and one of the algorithms tested separated the image from the plate and its characters adequately, however the character identifier produced results which were below the expected threshold. The likely cause for that is that Tesseract needs to be trained in the specific font used in the plates after image processing has been carried out.