基于颜色和纹理特征融合的马铃薯分级方法

目前马铃薯分级主要依靠人工完成,而对马铃薯的自动分级研究主要是对形状和大小进行初步分级,为了更好地对马铃薯进行精细化分级,提出一种颜色和纹理特征融合的分级方法,将马铃薯按色泽、损伤和蔫坏程度进行分级。首先提取马铃薯图像的马铃薯的HSV(hue saturation value)颜色特征,采用小波变换提取纹理特征,对两特征进行特征融合,然后通过对比支持向量机(support vector machine,SVM)分类器采用不同核函数的分级精度,选择了分级精度较高的径向基(radial basis function,RBF)核函数,最后模拟生产线上的马铃薯分级环境进行试验。试验结果为:能够将马铃薯按色泽、损伤和蔫坏程度分为三个等级,分级准确率达到97. 67%,每帧图像的平均处理时间为1. 0 s。该方法可用于马铃薯的精确分级,有利于提升马铃薯的质量和商品化速度。

Potato classification method based on fusion of color and texture features

At present, the classification of potato mainly depends on manual work, while the automatic classification of potato mainly focuses on the preliminary classification of shape and size. In order to better refine the classification of potato, a classification method based on the fusion of color and texture features is proposed, in which the potato is graded according to its color, damage and exhaustion degree. First extract HSV color characteristics of potato, potato using wavelet transform to extract the texture feature, feature fusion, the two characteristics and through comparing the SVM classifier using the classification accuracy of different kernel functions, chose the classification precision of RBF kernel function, finally simulated conditions of production line of potato classification experiment. The experimental results show that the potato can be divided into three grades according to color, damage and exhaustion, the classification accuracy is up to 97.67%, and the average processing time of each frame is 1.0s. This method can be used for the accurate classification of potatoes, which is beneficial to improve the quality and commercialization speed of potatoes.