SciELO - Scientific Electronic Library Online

 número49Aplicación móvil para monitoreo del consumo de materia seca en rumiantes usando la tecnología bluetoothAnálisis automático de micrografías SEM mediante aprendizaje profundo índice de autoresíndice de assuntosPesquisa de artigos
Home Pagelista alfabética de periódicos  

Serviços Personalizados




Links relacionados

  • Não possue artigos similaresSimilares em SciELO


RISTI - Revista Ibérica de Sistemas e Tecnologias de Informação

versão impressa ISSN 1646-9895


HUALLPA, Elias Ccoto; MACALUPU, Abel Angel Sullon; LUQUE, Jorge Eddy Otazu  e  SANCHEZ-GARCES, Jorge. Determination of the best license plate detection algorithm in controlled and uncontrolled environments. RISTI [online]. 2023, n.49, pp.83-99.  Epub 31-Mar-2023. ISSN 1646-9895.

Security is a priority in management and part of this is control and monitoring to detect situations that attack the public or private good. In this sense, the recognition of car license plates is added to these monitoring and control systems. This research work applied a series of artificial intelligence algorithms to automate such detection. In this sense, image processing functions were used with the OpenCV framework, considering that the information sources could have different scenarios, since the detection environment is open and exposed to the weather conditions of the area. The photos were taken in an environment located in southern Peru, whose conditions were rainy, sunny, and the day the hail fell. The image database was then divided into two categories; controlled environments where the same distance, a single angle, but not necessarily the same climate was considered; and uncontrolled environments with different angles, different distances and climates. Morphological transformation, Gaussian smoothing and Gaussian thresholding were also used for image processing. With the segmented images and the number plate digits extracted; 3 algorithms K-NN, SVM and Tesseract were compared. In each algorithm, its hyperparameters were used for the respective recognition of characters in the images, the following results were obtained with images taken with different angles and in different luminosities (uncontrolled environment) where the best Overall accuracy with 86% and the second group were images taken with a similar angle and similar distances (controlled environment), obtained an Overall accuracy of 95.5%.

Palavras-chave : Tesseract; OpenCV; Machine Learning and hyperparameters.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )