基于填充函数法训练BP神经网络的车牌字符识别算法  

An Improved License Plate Recognition Algorithm for Training the BP Neural Network Based on the Filled Function Method

作　　者：徐应涛[1] 陆福宏[2] 张莹[1,3] 

机构地区：[1]浙江师范大学数理与信息工程学院,浙江金华321004 [2]金华市公安局,浙江金华321004 [3]上海大学理学院,上海200444

出　　处：《计算机工程与科学》2009年第5期59-61,共3页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(10571116)

摘　　要：字符识别是车牌识别系统的一个关键问题。常用方法收敛速度慢,易陷入局部最优,用全局优化填充函数法训练BP神经网络的车牌字符识别算法可以跳出当前局部极小点,得到一个更低的极小点,重复此过程得到全局极小点,从而提高算法全局寻优能力。实验表明,该算法具有识别率高、识别速度快、车牌定位准确的特点,取得良好的运行效果。Character recogition is an important problem in license plate recognition systems. Aiming at the handicaps m the current methods such as slow convergence or easiness of getting into local optimization, this paper works out an improved license plate recognition algorithm based on the filled function method for training the BP neural network. This method can find a lower local minimizer by leaving the local minimizer previously found. By repeating these processes, a global minimizer can be obtained. Therefore, it can improve the ability of finding the global solution. The license plate recognition algorithm has a high recognition rate, a quick recognition speed, a precise license plate location and has been put into practice well.

关 键 词：车牌识别 字符识别 填充函数 BP神经网络 智能交通监控集成系统 

分 类 号：O221[理学—运筹学与控制论] TP391.4[理学—数学]