基于机器视觉的花生种子外观品质检测与分类方法研究  

Research on the Appearance Quality Detection and Classification of Peanut Seeds Based on Machine Vision

作　　者：马佳佳 王克强[1] 郑奕雄[1] 蔡肯[1] 林钦永 MA Jia-jia;WANG Ke-qiang;ZHENG Yi-xiong(Zhongkai University of Agriculture and Engineering,Guangzhou,Guangdong 510225)

机构地区：[1]仲恺农业工程学院,广东广州510225

出　　处：《安徽农业科学》2021年第10期225-227,231,共4页Journal of Anhui Agricultural Sciences

基　　金：广东省科技计划项目(2017A040405056);广东省研究生教育创新计划项目(KA200192346)。

摘　　要：[目的]花生种子的有效分类是花生产业中选育良种的重要环节,为降低花生产业对人工的依赖程度,简化选种过程,提出了一种基于机器视觉的花生种子外观品质检测与分类方法。[方法]在相同环境下采集单粒花生种子图像,建立花生种子对象在图像中像素数与其实际质量的回归模型,以花生种子尺寸和外观颜色作为主要特征,采用支持向量机分类模型完成分类任务。[结果]使用该方法完成12个类别的分类,对批量花生种子的分类准确率达86%,符合实际生产中花生种子初步分类要求。[结论]该方法对花生种子图像代表性特征的选择和识别样本的简化使得分类系统更符合实际生产需要,对同品种花生种子的不同品质分类以及不同品种花生种子的直接分类有着积极意义。[Objective]The effective classification of peanut seeds was an important link in peanut breeding.In order to reduce the dependence of peanut industry on labor and simplify the seed selection process,a method of peanut seed appearance quality detection and classification based on machine vision was proposed.[Method]The image of single peanut seed was collected in the same environment,and the regression model between the image prime number and the actual quality of peanut seed object was established.The classification task was completed by using support vector machine(SVM)classification model with the main characteristics of peanut seed size and appearance color.[Result]The classification accuracy of 12 categories was 86%,which met the preliminary classification requirements of peanut seeds in actual production.[Conclusion]The selection of representative features of peanut seed image and the simplification of recognition samples make the classification system more in line with the actual production needs.It has positive significance for different quality classification of the same peanut seed and direct classification of different peanut seeds.

关 键 词：花生种子 机器视觉 图像识别 支持向量机 

分 类 号：S126[农业科学—农业基础科学]