结合IGA和BP算法的物体识别方法  

Objects recognition based on improved genetic algorithm and BP algorithms

作　　者：刘勇[1] 张蕾[1] 范庆辉[1] 

机构地区：[1]河南科技大学电子信息工程学院,河南洛阳471003

出　　处：《计算机工程与应用》2009年第18期205-208,共4页Computer Engineering and Applications

基　　金：国家自然科学基金No.60475021;河南省杰出青年基金(No.0412000400)~~

摘　　要：目前常用的物体识别方法,其过程非常复杂,信息量和计算量都很大。结合改进遗传算法的神经网络方法,采用将结构与误差结合的适应度函数,改进的遗传算子实现对BP网络结构和权值的同步优化。提出一种用改进遗传算法优化后的BP神经网络进行物体识别,并以提取的修正不变矩特征作为BP神经网络的输入,仿真结果表明该方法提高了识别的稳定性和收敛性能,并且识别率较高。从而验证了该方法的有效性。Algorithms for recognition are cnmplcx,and the information and computation are large at present.Neural network based on improved genetic algorithm adopts fitness function of combining structure and error and the improved genetic operator to implement the optimization of structure and weights of BP network simultaneity.To recognize objections,BP neural network based on improved genetic algorithm is proposed in the paper,and the improved invariant moments extracted are regarded as the imput of BP network.The simulation results indicate that the method improves the stability and convergence capability of recognition.Moreover,the recognition rate is very high.So the efficiency is proved in the paper.

关 键 词：神经网络 遗传算法 特征提取 物体识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]