车牌倾斜校正研究综述

作为智能交通系统的核心技术之一,基于机器视觉的车牌识别一直受到广泛的关注.对车牌图像进行倾斜校正是车牌识别的重要步骤,其目的是解决拍摄角度随机性对识别过程的影响.该文将现有倾斜校正方法划分为直线检测、投影最值、角点检测、主方向分析四类,首先介绍了它们的研究究果,然后剖析了它们在鲁棒性方面存在不足的原因,展望了下一步的研究方向.

A Survey of License Plate Tilt Correction

License plate recognition based on machine vision, one of the key technologies of intelligent transportation system, has been paid much attention to. Tilt correction for license plate image plays an important role in license plate recognition and its aim is to weaken the influence of capture angle randomness to recognition process. In this paper, tilt correction methods are divided into four types： line detection, projection maximum or minimum, corner detection and principal orientation analysis. After introducing research achievements of these methods, we analyze the causes of shortages related their robustness and look into the development trends of tilt correction.