PROJECT TITLE :

A Robust and Efficient Approach to License Plate Detection - 2017

ABSTRACT:

This paper presents a sturdy and economical methodology for car place detection with the purpose of accurately localizing vehicle license plates from advanced scenes in real time. A straightforward yet effective image downscaling methodology is 1st proposed to substantially accelerate registration number plate localization without sacrificing detection performance compared with that achieved using the original image. Furthermore, a completely unique line density filter approach is proposed to extract candidate regions, thereby considerably reducing the world to be analyzed for vehicle plate localization. Moreover, a cascaded vehicle plate classifier based mostly on linear support vector machines using color saliency features is introduced to identify the true vehicle plate from among the candidate regions. For performance analysis, a knowledge set consisting of 3977 images captured from numerous scenes under completely different conditions is also presented. Extensive experiments on the widely used Caltech license plate data set and our newly introduced information set demonstrate that the proposed approach substantially outperforms state-of-the-art methods in terms of each detection accuracy and run-time potency, increasing the detection ratio from ninety one.09% to 96.sixty two% whereas decreasing the run time from 672 to 42 ms for processing a picture with a resolution of 1082×728. The executable code and our collected knowledge set are publicly obtainable.