COMPARISON OF METHODS FOR AUTOMATIC LICENSE NUMBER RECOGNITION

The paper is devoted to the problem of automatic detection and recognition license plates, solution which has many potential applications, from security traffic management. purpose this work was compare methods finding recognizing car number based on application deep learning algorithms, takes into account different regional standards video quality, speeds vehicles, location camera in relation vehicle plate, defects plate (pollution , deformation), as well changes external lighting conditions. advantages disadvantages localization segmentation plates cars using image binarization, Viola–Jones Harr are given. It determined that adaptive approaches better due possibility compensating impact obstacles areas image, for example, distribution shadows heterogeneity illumination. real algorithms rely directly or indirectly presence limits. Even if limits not used when determined, they have be further analysis. templates, histograms, contour analysis were compared identify familiar features (segmentation). shown an effective approach can Viola-Jones, Harr, brightness histograms SVM method. Formulated conclusions effectiveness implementation each procedures confirmed a result conducting experiments with developed software python 3 language cv2 computer vision library. described makes it possible obtain fairly high accuracy at angles rotation relative camera. Keywords: recognition, localization, normalization, segmentation, character recognition.