基于特征显著性的多特征融合车牌定位算法

在分析已有车牌定位技术以及目标检测共有特性的基础上,提出了基于视觉显著性的特征选择方法.算法依据先验样本的统计学习,利用导致最小错误概率判决方法,得到目标的特征显著性分析.在车牌定位过程中,根据得到的特征显著性序列,依次赋予特征不同的权值,然后采用融合的方式得到所需要的车牌区域.实验结果表明,该算法提高了使用单一特征进行车牌定位的准确率.

Multi-features fusion license plates locating algorithm based on featuresalience

Based on the analysis of license plates technique and the commonness of target detection, the feature selection
method based on vision salience is proposed. According to prior-sample training and minimum probability of error, feature
salience is obtained. In real license plate location, the more salient features are given the larger weights. Then, the system
fuses the multi-features to locate license plates. Experimental results show that the proposed method has better identify rate
than single feature.