基于统计纹理的柑橘糖度与有效酸度检测  

Determination of Sugar Content and Valid Acidity of Citrus Fruit Based on Statistics Texture

作　　者：曹乐平[1] 温芝元[2] 陈理渊[3] 

机构地区：[1]湖南生物机电职业技术学院现代管理工程系,湖南长沙410127 [2]湖南农业大学理学院,湖南长沙410128 [3]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《测试技术学报》2009年第1期63-67,共5页Journal of Test and Measurement Technology

基　　金：湖南省自然科学基金资助项目(2007JJ6129);湖南省农业厅重点课题(200704A);湖南省教育厅科学研究项目(06D059)

摘　　要：用柑橘图像的纹理统计量无损检测柑橘的糖度和有效酸度.对柑橘图像蓝色分量设置阈值,去除图像背景,修正去背景后柑橘图像灰度直方图,提取灰度图像柑橘区域熵、三阶矩、均值和标准差4个纹理统计量,建立温州蜜柑糖度和有效酸度偏最小二乘预测模型.30个样本试验结果表明:±1°Brix范围内糖度判断正确率为76.67%;±0.5 pH及±0.25 pH范围内有效酸度判断正确率分别为100%和80%.比较2个预测模型中纹理统计量间权值的大小发现:熵的权重最大.这说明:柑橘果皮纹理与柑橘的糖度及有效酸度间存在对应关系,且这种对应关系中熵是关键指标.Non-destructive detection of sugar content and valid acidity is crucial for online integrative grading of fruit exterior/interior quality. Blue component threshold is used to filter the background of citrus fruit images, and then the image grayscale histogram is modified. Texture statistics as regional entropy, third-order moment, mean value and standard deviation of grayscale images are extracted. Predictive models of sugar content and valid acidity of Wenzhou citrus fruits are constructed with partial least square method. Results of 30 test samples show the correctness for accuracy ±0.1° Brix of sugar content is 76. 67%, for valid acidity, correctness for accuracy ±0. 5 pH and ±0.25 pH is 100% and 80% respectively. Comparing weights of texture statistics in two predictive models, it is found that the weight of entropy is significant big, which means that citrus fruit texture correlates to sugar content and valid acidity, and entropy is the critical parameter in predictive models.

关 键 词：纹理统计 柑橘 偏最小二乘法 糖度 有效酸度 

分 类 号：TP319.41[自动化与计算机技术—计算机软件与理论]