基于多光谱特征的柑桔品质检测算法研究  

Quality Detecting Algorithms of Orange Based on Multi-spectral Imaging

作　　者：李翠[1] 严治国[1] 朱均超[1] 杨进风[1] 

机构地区：[1]天津理工大学天津市复杂系统控制理论及应用重点实验室,天津300384

出　　处：《计算机仿真》2012年第9期262-265,共4页Computer Simulation

基　　金：大学生创新性实验计划项目(091006015);新疆维吾尔自治区科技计划项目(201154124)

摘　　要：为解决柑桔自动检测和分级中的品质缺陷检测问题,针对检测速度慢和正确率不高,提出一种基于多光谱技术新的柑桔品质检测算法。检测系统主要由光照箱、多光谱光源、摄像机、图像采集卡及计算机等组成。采用最小二乘支持向量机(LS-SVM)法将柑桔的形状等特征提取出来,并提出了基于可调多光谱光源图像采集方式,利用单目视觉技术和多目视觉技术相融合的方法进行水果表面缺陷的检测。实验结果表明,多光谱图像技术能够快速有效地检测水果表面缺陷,对柑桔缺陷检测正确率达到97%。To solve the detection problem of orange defect in the automatic detection and classification, a new al-gorithm based on multi-spectral Imaging was developed. Detection system consists of light boxes, multi-spectrum light source, camera, image acquisition and computer etc. Using LS-SVM to extract orange size, shape and other characteristics, and a method of image acquisition was proposed, based on tunable multi-spectral light source. The defective parts of oranges can be extracted effectively by the integration of monocular vision and multi-spectral vision method. The experimental results show that the multi-spectral imaging technology can detect fruit surface defect quickly and effectively and its correct rate is up to 97%.

关 键 词：多光谱图像 最小二乘支持向量机 技术融合 品质检测 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]