基于PSO-BP神经网络的车牌号码识别技术  

License plate recognition technology based on PSO-BP neural network

作　　者：吴志攀[1] 赵跃龙[2] 罗中良[3] 杜华英 WU Zhipan ZHAO Yuelong LUO Zhongliang DU Huaying(School of Information Science and Technology, Huizhou University, Huizhou 516007, China School of Computer Science and Engineering, South China University of Technology, Guangzhou 510640, China School of Electronic Information and Electrical Engineering, Huizhou University, Huizhou 516007, China Department of Information Technology, City College of Huizhou, Huizhou 516025, China)

机构地区：[1]惠州学院信息科学技术学院,广东惠州516007 [2]华南理工大学计算机科学与工程学院,广东广州510640 [3]惠州学院电子信息与电气工程学院,广东惠州516007 [4]惠州城市职业学院信息技术系,广东惠州516025

出　　处：《中山大学学报（自然科学版）》2017年第1期46-52,共7页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：国家自然科学基金(61572200);广东省高等学校教学质量与改革工程本科类项目([2013]113号-113);惠州市科技计划项目(2014-01);惠州城市职业学院课题(HZC2015-5-1-3)

摘　　要：针对传统BP神经网络在车牌识别的应用领域中,存在经常性陷入局部最优,而导致识别效果不理想事实。提出一种基于PSO-BP神经网络的车牌号码识别技术方法,该方法首先构建一个8-25-1的BP神经网络,再通过提取车牌的8像素比特征值作为BP神经网络的输入向量,然后采用粒子群算法(PSO)对该BP神经网络的权值和阈值进行优化,使其适应值达到最小。通过300副不同光照环境和污损的车牌识别仿真实验,验证了所提出的算法相对于传统的模板匹配算法和BP算法,具有输出误差小、全局收敛速度快和识别率高的特征,具有一定的应用价值。BP neural network has been successfully applied in many fields,such as license plate recog-nition,however the general method of BP neural network is likely to fall into local optimum rather than converge to the global optimum,so its performance of recognition is not very satisfied.In order to avoid this situation,a license plate number recognition technology method is proposed based on PSO-BP neural network.The key of this method is to build an 8-25-1 BP neural network with 8 pixel rate features as its input vector firstly,and then use particle swarm optimization （PSO）algorithm to optimize the weights and thresholds of the BP neural network to get the best fitness value of global searching.By using 300 pair license plate images some with the different illumination and the stained parts,the simulation result indicates that the algorithm we mentioned comparing with the traditional template matching algorithm and general BP algorithm has the certain application value with the advantage of less output errors,global fast convergence,and high recognition rate.

关 键 词：BP神经网络 粒子群算法(PSO) PSO—BP神经网络 车牌识别 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]