基于PSO-RBF神经网络的雾霾车牌识别算法研究

The Haze Plate Recognition System Based on PSO-RBF Neural Network

给出一种雾霾环境下车牌识别改进算法.首先利用改进的暗原色先验法对雾霾天气下的车牌图像进行去雾处理;然后经预处理、定位、分割与提取,得到粗网格特征矩阵;最后采用经粒子群算法优化的径向基函数神经网络进行识别.实验结果表明,系统去雾效果良好,且能缩短去雾处理的时间,有效提高雾霾天气下车牌识别的速度和准确率.

In this paper, a new algorithm of license plate recognition in the hazy weather was designed. First- ly, defogging operation was introduced for license plate image in the environment of hazy by using improved dark channel prior. Then after the pretreatment, positioning, segmentation and extraction, coarse grid charac- teristic matrix is obtained. Finally, radial basis function （RBF） neural network, which was optimized by par- ticle swarm algorithm in advance, was used to identify the character. The experiment results showed that the improved algorithm not only had a good effect on haze removal, but also reduced the duration of defogging, which effectively improve the license plate recognition speed and accuracy in fog and haze weather.