摘要
在油茶果脱壳后,采用机械方法分选的茶籽中混杂着一些果壳,由于两者大小和比重相似,其外观差异成为分选的重要依据。本文提出一种综合考虑油茶果颜色、纹理、几何形状多种图像特征的分选方法,结合机器视觉技术实现油茶果果壳与茶籽的准确分选。通过有效的图像预处理手段提取物料样本轮廓,并计算轮廓内的颜色、纹理以及形状特征信息,分别比较了通过网格搜索法(Grid Search,GS)、粒子群算法(Particle Swarm Optimization,PSO)和遗传算法(Genetic Algorithm,GA)寻优而建立的三种支持向量机(Support Vector Machine,SVM)分类模型,最终确定GS-SVM模型最佳,其模型训练集识别率为94.44%,测试集识别率为93.33%,结果表明将此方法应用于油茶果果壳与茶籽分选是可行的,为油茶果果壳与茶籽分选加工技术提供了一定的理论基础。
After shelling the camellia,camellia seeds which sorted by mechanical means,was mixed with some shells.Because of the similar size and specific gravity between the two,the appearance difference became the important sorting basis.A method that comprehensively considered multiple image features of camellia,such as color,texture and geometric shape is proposed,which combined with machine vision technology to achieve accurate sorting for shells and seeds.The contour of material was extracted by effective image preprocessing method,and the information of color,texture and geometric shape within the contour were calculated.Three kinds of support Vector Machine classification models established by grid search,particle swarm optimization and genetic algorithm were compared respectively,and the GS-SVM model was determined to be the best finally.The recognition rate of model training set was 94.44%and 93.33%respectively.This paper provided a certain theory for the sorting method of the seeds and shells of the camellia.
作者
段宇飞
皇甫思思
王焱清
汤旸
霍俊
Duan Yufei;Huangfu Sisi;Wang Yanqing;Tang Yang;Huo Jun(Institute of Agricultural Machinery,Hubei University of Technology,Wuhan,430068,China;Hubei Province Engineering Technology Research Center for Intelligent Agricultural Machinery,Wuhan,430068,China)
出处
《中国农机化学报》
北大核心
2020年第6期171-178,共8页
Journal of Chinese Agricultural Mechanization
基金
湖北工业大学科研启动基金项目(BSQD2017076)
湖北省重点研发计划项目(2015BBA211)。
关键词
机器视觉
油茶果
分选
支持向量机
图像特征
machine vision
camellia
sorting
support vector machine
image feature
shells
seeds
