摘要
论文提出了一种基于EMD进化概率神经网络的纹理图像识别方法。首先,对原始信号进行经验模式分解,将其分解为多个平稳的固有模式函数之和;再从各IMF分量中提取主要能量特征作为进化概率神经网络的输入参数来识别纹理图像。对不同的自然纹理图像进行了实验,并将结果与小波进化概率神经网络的结果做了比较。实验结果证明,论文方法的正确识别率和识别精度高于小波进化概率神经网络。
A new texture image recognition method based on empirical mode decomposition and differentia evolution probabilistic neural network is proposed in this paper. At first, It takes the empirical mode decomposition on the original sig- nal and decomposes it into the sum of multi steady Intrinsic Mode Function(IMF) ; Then the mainly energy features are taken as evolution probabilistic neural network's input parameter from every IMF to identify the texture image. Experiments are conducted on natural different texture images. Compared to wavelet-evolution prohabilistic neural network, the method proposed in this paper has the higher correct recognition rate and identification accuracy.
出处
《计算机与数字工程》
2014年第9期1713-1716,共4页
Computer & Digital Engineering
基金
陕西省教育厅2014年科学研究计划专项项目(编号:14JK2037)资助.
关键词
经验模式分解
固有模式函数
概率神经网络
差异进化
纹理分类
empirical mode decomposition, IMF, probabilistic neural network, differentia evolution, texture classifi-cation