Abstract: Potato shape is one of the important indicators of potato grading, this paper discusses a potato shape classification method based on machine vision technology, combined with the principal component analysis-support vector machine (PCA-SVM) algorithm. This method extracts the eigenvectors of eleven-dimensional in a single potato region，which can represent shapes, and principal component analysis (PCA) is used to reduce the dimension of the feature vector and extract the principal component features of the shape. Then, the principal component feature is brought into the support vector machine(SVM)for modeling, and the grid search method (GS) is used to optimize the parameters of SVM. In the detection, the images of potato samples were taken into the PCA model and the optimized SVM model by using the ten-fold cross validation (CV) algorithm to classify the potatoes. Experimental results show that the algorithm proposed in this paper has the higher sorting speed as well as the higher accuracy (97.3 %). The method is feasible for potato shape sorting and can be used for automatic potato grading.
Key words: Potato, machine vision, shape sorting, principal component

摘要： 马铃薯形状是马铃薯分级的重要指标之一,本文利用机器视觉技术根据形状对马铃薯进行分选。提出一种结合主成分分析-支持向量机(PCA-SVM)算法的马铃薯形状分选方法。该方法提取单个马铃薯区域内11维表征形状的特征向量，并利用主成分分析(PCA)法对特征向量进行降维，提取出形状的主成分特征，然后，将主成分特征带入支持向量机(SVM)进行建模，并利用网格搜索法(GS)对SVM进行参数优化。检测中，采用十折交叉验证(CV)算法依次将马铃薯样品图片带入PCA模型和参数优化好的SVM模型，实现对马铃薯的分类。实验结果显示，提出的算法具有较强的可行性，分选速度快且准确率达97.3 %，能用于马铃薯自动化分级。

关键词: 马铃薯, 机器视觉, 形状分选, 主成分分析, 支持向量机, 网格搜索