Super-Resolved Recognition of License Plate Characters

Object detection and recognition are crucial in the field of computer vision an active area research. However, actual object processes, accuracy is often degraded due to resolution mismatches between training test image data. To solve this problem, we designed developed integrated super-resolution framework by proposing technique that improves accuracy. In detail, collected a number license plate images through web-crawling artificial data generation, neural network was trained defining objective function be robust flips. verify performance proposed algorithm, experimented with on representative confirmed character recognition. For 4× magnification, method remarkably increased mean average precision 49.94% compared existing state-of-the-art method.