Modelo de reconocimiento automático y detección de matrículas basado en OpenCV y Machine Learning

Automatic number plate recognition (ALPR) is an important task with many applications in intelligent transportation and surveillance systems. This research takes into consideration the functions of image processing for the detection and recognition of number plates; which can come from noisy sources, low illumination, different angles and distances taken from the images (uncontrolled environments); most of the existing automated number plate recognition systems only work in a controlled environment where images are captured from a right angle with good illumination and clarity [25]. According to [11] in uncontrolled environments the probability of recognising the characters on the plates decreases and this has been observed in the research. To achieve image processing, morphological transformation, Gaussian smoothing and Gaussian thresholding were used, then 3 different algorithms K-NN, SVM and Tesseract were used for character recognition, each algorithm with their respective hyperparameters for optimisation. The images were separated into two groups, the first with 80 images taken from different angles and distance (uncontrolled environment) where the best Overall accuracy was obtained with 86 % and the second group were images taken at a right angle and similar distance (controlled environment), this group obtained an Overall accuracy of 95.5 %.