依据彩色共生矩阵对木质板材的识别

Wood Panel Recognition by Color Co-occurrence Matrix

为了实现对木质板材种类识别智能化,提出一种基于彩色灰度共生矩阵的识别方法。将原彩色图像色调、饱和度、明度颜色空间3个通道的图像作为伪灰度图像,分别获取其灰度共生矩阵纹理特征参数,组成一个42维特征向量。使用遗传算法进行特征选择后,得到优化后的22维特征向量,在此特征向量形成的特征空间中,概率神经网络(PNN)分类器的识别率为96.0%。试验结果表明,此方法用于解决木质板材分类识别的问题是可行的。

In order to achieve intelligent identification of the wood panel species,a method with color co-occurrence matrix was proposed.Three color channels’image(H,S,V)of the original color image were took as pseudo gray image,and then gray level co-occurrence matrix feature parameters were obtained,which was a 42 dimension feature vector.The optimized feature vector was obtained by the feature selection method based on genetic algorithm,consisting of 22 dimension feature parameters.With the optimized feature vector and PNN neural network classifier,the recognition rate reached to 96.0%.The result indicates that the method proposed can better solve the wood panel classification and recognition.