一种车牌识别系统中的混合车牌定位方法

提出了一种新的粗定位和精细定位相结合的车牌定位方法。首先采用一种基于垂直边缘和车牌句法的车牌粗定位算法，确定车牌候选区域，剔除伪区域，粗定位可以快速地将
车牌从原始车辆图像中截取出来，同时精确确定车牌的上下边界。然后再进行基于旋转差分投影的车牌水平精确定位与倾斜校正，由于水平定位和倾斜校正同时进行，提高了水平分割的精度并降低了定位时间。实验表明该混合方法能快速、精确地进行车牌定位，有效地解决了复杂背景下的车牌定位精度不高的问题，同时对模糊车牌具有很强的抗干扰性。
关键词：粗定位；伪区域；车牌句法；车牌定位；
Abstract: A new hybrid method for license plate position is proposed based on rough position and precision position. Firstly, a rough position algorithm based on rough edge and plate syntax is adopted to decide the candidate region and discard pseudo-region, which can find the plate image from the original vehicle image rapidly and decide its upper and lower borders. Secondly, level accurate position and incline correction are done on the basis of rotating difference projection. Because of their concurrence, level segmentation accuracy is improved and positioning time is reduced. The experimental research shows that the hybrid position method can be used to solve some problems such as lower accuracy in complicated scenes in terms of its rapid speed, high accuracy, good self- adaptivity, and anti-disturbance.
Keywords: rough location；pseudo-region; plate syntax; plate position;