欢迎访问《图学学报》
分享到：
摘 要：在分析人耳Gabor特征基础上，提出一种主成分分析降维并利用基于粒子群优化训练的人工神经网络对部分遮挡人耳进行识别方法。选取了PCA方法降维后人耳图像的Gabor特征值作为人工神经网络训练样本，利用粒子群优化算法与多层前馈网络结合算法训练神经网络。与多种方法对比的实验表明，针对部分遮挡人耳的测试实验，基于Gabor+PCA特征与粒子群算法的部分遮挡人耳识别方法具有高识别性能，取得好的效果。#br# 关 键 词：人耳识别；Gabor小波特征；主成分分析；粒子群优化算法
Abstract: On the basis of the human ear Gabor features, PCA dimensionality reduction and the artificial neural network based on PSO training are used to identify partially occluded human ear. The ear image PCA dimensionality reduction of Gabor feature value is selected as the training samples of neural network. The network is trained to take the particle swarm optimization algorithm combined with BP algorithm. Comparative experiments with other methods indicate that the partially occluded ear recognition based on the Gabor wavelet transform + principle component analysis and PSO-BP neural network method has higher recognition performance and can achieve good results.#br# Key words: ear recognition; Gabor feature; principle component analysis; particle swarm optimization
摘要： 在分析人耳Gabor特征基础上，提出一种主成分分析降维并利用基于粒子群优化训练的人工神经网络对部分遮挡人耳进行识别方法。选取了PCA方法降维后人耳图像的Gabor特征值作为人工神经网络训练样本，利用粒子群优化算法与多层前馈网络结合算法训练神经网络。与多种方法对比的实验表明，针对部分遮挡人耳的测试实验，基于Gabor+PCA特征与粒子群算法的部分遮挡人耳识别方法具有高识别性能，取得好的效果。
