基于差异演化概率神经网络的纹理图像识别.pdfVIP

Computer and Applications 计算机工程与股用 2∞8 ， 44 (11) 179 基于差异演化概率神经网络的纹理图像识别 杨鼎强，肖淑苹，蒋加伏 YANG Ding-qiang,XIAO Shu-ping,JIANG Jia-fu 长沙理工大学计算机与通信工程学院，长沙 410076 College of Computer and Communication Engineeri吨， Changsha University of Science and Technology. Changsha 41 ∞76.China E叫 mail: ydq966@126.com YANG Ding-qiang,XIAO Shu叩ping ， JIANG Jia阳仇I.Texture image reco朗副ion based on Differentia Evolution probabilistic neural network.Computer Engineering and Applica创ons ， 2008 ， 44( 11): 179-181. Abstract:ηle Differential Evolution (DE) method is introduced in thís paperωmake up the shortage of basic probabilistic neural network.Consequently, a new texture image reco伊 ition method based on Differentia Evolutíon probabilistic neural network is proposed.At fi酬， it extracts the energy character with the shape of tree stru盯 ture wavelet packet transform and extracts the statistical mean value , average energy. standard deviatíon , mean resídual characteristics with the statistic method , the feature vector ís obtained by the above characteristics.Then the feature vector of the texture ímage is trained by the Differentia Evolutíon prob? abilistic neural network , thus the texture classification is identified. The experiment result indícates: compare 协 the BP neural net? work. RBF neural network and the basic probabílistic neural network , the Differentía Evolutíon probabilistic neural network has the higher accuracy and the faster convergence speed. Key words: texture classíficatíon;wavelet packet transform;probabílistic neural network;Differentia Evolution(DE) 摘要 :~I 入差异演化 (DE) 算法来弥补基本概卒神经网络的不足，从而提出一种基于最异演化概率神经网络的纹理图像识别方 法。首先用树形结构小波包交换提取纹理图像的能受特征，用!.于统计的纹理特征方法提取统计均值、平均能量、标准浆和平均残 余特征，得到纹攻图像的特征矢量;然后用差异演化概率神经网络训练纹理图像的特征矢量，从而实现纹理图像的识别。实验结果 农明:该方法较 BP 神经网络、 RBF 神经网络和基本的 PNN 有更高的识别正确卒，