摘要
研究车牌自动识别系统中的车牌特征选择问题。车牌特征选择结果的优劣将直接影响车牌识别准确率,因此选择对车牌字符识别能力强的特征是车牌自动识别中一个关键问题。由于传统特征选择方法计算时间长及标准遗传算法过早收敛的缺点,导致车牌字符识别准确率低。为了提高车牌识别的准确率,提出一种改进遗传算法的车牌特征选择方法,把车牌特征作为遗传算的种群中个体,以车牌识别率为目标函数,在算法前期采用保优策略,保持最优特征的多样性,来避免局部最优;在进化后期,当特征最优信息陷入停滞时,自适应调整进化参数,加快进化速度。仿真结果表明,选择车牌特征的车牌自动识别系统识别准确率达到了92.36%,远远高于标准遗传算法和自适应遗传算法的86.53%和90.3%,说明,改进方法不仅提高了车牌字符识别率,而且克服了传统遗传算法在特征选择过程中出现的早熟现象,防止陷入局部最优的缺点。
In the automatic recognition system of vehicle license,feature selection for vehicle character recognition is the key factor in pattern recognition.Aimed at low precision and over early convergence problems,the paper proposed a new intelligent genetic algorithm(IGA) and thereby proposed a new method for feature selection.IGA uses select strategies to ensure the quality and adaptive crossover and mutation operators,and improve the vehicle plate recognition performance.The results of the simulation show that the intelligence genetic algorithm not only is better than the SGA and AGA in respect of convergence speed and search ability,but also effectively avoids precocity and plunging into local optimum.IGA algorithm improves the license plate recognition,and the recognition speed is greatly accelerated.
出处
《计算机仿真》
CSCD
北大核心
2010年第12期331-334,344,共5页
Computer Simulation
关键词
车牌识别系统
特征选择
标准遗传算法
Vehicle license recognition system
Feature selection
Simple genetic algorithm