基于颜色特征对木质板材分级的研究

Research on Wooden Board Classification Based on Color and Texture Features

给出几种基于颜色和纹理特征木质板材的分级算法,这几种方法通过主颜色和共生矩阵特征,分别利用神经网络中的RBF网络和K-近邻对木质板材样本图像进行处理,然后进行分类的方法。木质板材样本图像通过对木材加工处理采集而得到的。主颜色特征是反映一幅图像的基本面貌的,并通过HSV颜色模型、量化、颜色直方图和绝对距离而得到的;再加上共生矩阵特征。选取三种主颜色方案数据,经过Matlab7.0平台进行程序设计,对木质板材样本图像数据进行训练和测试,并加以验证。实验结果表明,这几种方法能够较好地解决木材分级的问题。

Several graduation algorithms for wooden board were given based on the color and texture feature, with which the specimen images of wooden board were processed by the RBF neural network and K- nearest neighbor separately based on the main color characteristic and co-occurrence matrix characteristic, and then the wooden boards were classified. All the original images of wooden board specimens were collected after the specimens were preprocessed. Here the main color characteristic which reflected the basic appearance of an image was obtained through the HSV color model, the quantification, the color histogram and absolute distance. With the main color and co-occurrence matrix characteristic, the data of three main color plans of the collected image data of wooden board specimens were trained and tested, then verified finally on Matlab7.0 platform. The result shows that all these methods can classify the wooden board well.