基于Tent映射CPSO和车牌纹理特征的车牌定位

License plate location based on texture features and tent chaotic particle swarm optimization

针对现有车牌定位算法定位准确率不高和速度慢等问题,结合车牌纹理特征,提出了一种基于Tent映射混沌粒子群(CPSO)的车牌精确定位算法.首先用基于二维直方图区域斜分的OTSU方法对车牌图像做二值化处理;接着使用三组一维滤波器获取其二值纹理特征向量.然后利用基于Tent映射CPSO快速准确的全局搜索能力,结合二值纹理特征向量构造适应度函数,并引入车牌纹理的一致性度量作为判决条件,找到车牌区域的最佳定位参量.最后,与基于遗传算法(GA)和基本粒子群算法(BPSO)的定位方法进行了比较.实验结果表明,该方法适应性强,定位效果较好,运行时间更短.

Considering the problems of the low precision ratio and slow arithmetic speed of license plate location, an accurate license plate location method based on tent chaotic particle swarm optimization （TCPSO） was proposed by combining the texture features. First, binarization was adopted to segment the license plate image by the OTSU method, which is based on a 2-D histogram oblique. Then the texture feature vector was obtained by three one-di- mensional filters. With the rapid and accurate searching ability, the best location parameters of license plate area were found by constructing the fitness function with the texture feature vector when introducing the texture coher- ence into the judgment. At last, the proposed method was compared with a genetic algorithm （GA） and BPSO. The experimental results show that the proposed method has stronger adaptability, better location effect, and shorter running time.