基于多特征降维的油茶果壳籽粒的分选识别  

Recognition and sorting of the shell and the seed for Camellia based on the reduction of multi-features dimension

作　　者：李昕 陈泽君[2] 李立君[3] 谭季秋[1] 吴发展 LI Xin;CHEN Zejun;LI Lijun;TAN Jiqiu;WU Fazhan(School of Mechanical Engineer Hunan Institute of Engineering,Xiangtan,Hunan 411104,China;Hunan Academy of Forestry,Changsha,Hunan 410004,China;School of Mechanical and Electrical Engineer,Center South University of Forestry and Technology,Changsha,Hunan 410004,China;Fengke Forestry Equipment Technology Co.Ltd,Zhuzhou,Hunan 412000,China)

机构地区：[1]湖南工程学院机械工程学院,湖南湘潭411104 [2]湖南省林业科学院,湖南长沙410004 [3]中南林业科技大学机电工程学院,湖南长沙410004 [4]株洲丰科林业装备科技有限责任公司,湖南株洲412000

出　　处：《湖南农业大学学报（自然科学版）》2022年第4期501-506,共6页Journal of Hunan Agricultural University(Natural Sciences)

基　　金：国家重点研发计划项目子课题(2016YFD0702100)。

摘　　要：针对目前株洲丰科林业装备有限公司智能–1型油茶脱壳机中分选识别系统存在的识别方法单一、受分选目标颜色影响大、自适应功能较差等问题,建立了基于多特征降维的油茶果壳籽粒的分选识别方法:提取油茶果壳籽粒的6维形态和颜色特征;根据这些特征的特性采用降维的方法,以保证算法的识别效率;将降维方法融入人工免疫网络算法中进行算法模型的辨识。选用颜色特征区分较明显的分别已晾晒3 d和12 d的油茶果进行采集分选,通过降维优化得到2分量、4分量与6分量的3 d晾晒样本识别率均值达70%、80%、90%;晾晒12 d的识别率均值达50%、60%、75%;晾晒3 d识别时间均值为60 ms、350 ms、450 ms;晾晒12 d的识别时间均值为80 ms、420 ms、480 ms。Aiming at the problem of single recognition method,large influence by target color and poor adaptive function for the recognition and sorting system of intelligent-1 type Camellia sheller manufactured by Fengke Forestry Equipment Technology Co.Ltd,a new recognition and sorting system method of the shell and the seed was established based on the reduction of multi-features dimension for Camellia.Six dimensional of morphological and color features was extracted in Camellia shell/seed.According to the characteristics of these features,a dimension reduction method is proposed to ensure the recognition efficiency of the algorithm.The dimension reduction method is integrated into the artificial immune network algorithm for multi-features dimension reduction identification.Camellia fruits of drying 3 d and 12 d with obvious color characteristics were selected for collection and sorting,respectively.Through dimension reduction optimization,the average recognition rate of 2-component,4-component and 6-component for drying 3 d samples reached 70%,80%and 90%,and 50%,60%and 75%for drying 12 d.The average recognition time was 60 ms,350 ms and 450 ms in 3 d,and 80 ms,420 ms and 480 ms in 12 d.

关 键 词：油茶果 分选 颜色特征 形态特征 多特征降维 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]