Abstract: To improve the efficiency of the spherical fruits and vegetables’ sorting and reduce the cost of sorting, a new spherical fruits and vegetables’ sorting method based on multiple feature combination under the machine vision technology is proposed. For single feature describe image is uncomprehensive, a combinative feature model of color moments, Zernike moments and gray level co-occurrence matrix is established, it is used to determine the comprehensive characteri-stics of fruits and vegetables. Before the calculation of Zernike moment , binary image edge extraction based on the H component threshold is carried out. Using BP neural network and support vector classifier for sorting the experimental samples respectively, by simulation experiments, the algorithm of multiple features combination’s feasibility and effectiveness is verified. The BP neural network and support vector machine classifier’s sorting effect is contrasted and analyzed, sorting rate both reaches more than 95%.

Key words: color moments, Zernike moments, gray level co-occurrence matrix, support vector machine, fruits and vegetables&rsquo, sorting

摘要： 为提高球形果蔬分选效率以及降低分选成本，提出了一种在机器视觉技术下球形果蔬多特征组合的智能分选方法。针对单一特征刻画图像特征不全面的问题，建立了颜色矩、Zernike矩、灰度共生矩阵三种特征的组合特征模型，用以确定果蔬的综合特征。Zernike矩在计算前进行了基于H分量阈值二值化图像边缘提取。利用BP神经网络和支持向量机构造分类器，分别对实验样本进行分选。通过仿真实验，验证了多特征组合算法的可行性和有效性，对比分析了BP神经网络和支持向量机分类器对分选效果的影响，分选率均达到了95%以上。

关键词: 颜色矩, Zernike矩, 灰度共生矩阵, 支持向量机, 果蔬分选