摘要
【目的】采用机器视觉技术对新疆冰糖心红富士苹果进行重量、糖度预测和分级。【方法】分析提取苹果RGB图像中单色、波长差、HSV转换后分量等多类型图像,对比图像分割效果确定后续处理图像。采用形态学处理剔除二值化图像果梗区域,提取目标区域几何、灰度和色调频度等特征。采用多元线性和偏最小二乘回归预测苹果重量和糖度,判别分析分类苹果,结合全组合实验方法和特征优选,获得较佳特征集合。【结果】多元线性回归方法建立苹果糖度的预测模型结果最佳,使用几何和灰度的特征集合,建模集和验证集糖度预测相关系数分别为0.623和0.570;使用面积、周长、长轴长度和短轴长度特征集和,或体积、周长、长轴长度和短轴长度四个特征时,多元线性回归预测苹果重量,验证集预测相关系数r为0.992,预测均方根误差为3.88 g,相对分析误差为8.1;采用基于特征优选方法确定41个主要特征,二次判别函数的判别分析分级苹果,验证集分级准确率达到98.7%。【结论】RGB图像能够准确预测新疆冰糖心红富士苹果重量,并能精确分级,但糖度预测效果不佳。
[ Objective ] The study aimed to investigate the prediction of the sugar degree, weight and the grading of Xinjiang Fuji apple by using RGB image. [ Method] First, different images were obtained from RGB image of the apple, including r, g and b monchrome images, h, s and v images of HSV transformed from RGB image, and wavelengths subtraction of r -b and g -b. After the segmentation were compared, the optimal image was selected for subsequent analysis. Second, results of these images the stem regions were removed from the original binarization images of apples by using morphological processing. Then, three features of geometry, gray and hue of segmentation regions were calculated from the stem - removed binearization images. Finally, the MLR and PLSR were used to predict the weight and sugar degree, and the DA by applying quadratic function for classifying apple grade. [ Result ] The results indicated that the correlation coefficients at the modeling set and validation set were 0.623 and 0. 570, respectively when the geometry, gray features were used to predict the sugar degree by MLR; When area, perimeter, major axis length , minor axis length and multiple linear regression were used to predict the weight, the correlation coefficient （r） and MECV were 0.992 and 3.88 g at the validation set, and the analysis error was 8.1 ; And the accuracy rate of apple grading was up to 98.7% when using quadratic discriminate analysis and 41 features of segmentation region from apple＇s three view, which is selected by wrapper method from 48 original features. [ Conclusion ] The weight and the grading of Xinjiang Fuji apple were accurately analyzed by using RGB image, while the prediction accuracy of the sugar degree was not high.
出处
《新疆农业科学》
CAS
CSCD
北大核心
2013年第3期509-517,共9页
Xinjiang Agricultural Sciences
基金
国家自然科学基金项目(61005022)
新疆维吾尔自治区科技厅基金项目(2009211B07)
关键词
机器视觉
红富士苹果
糖度
重量
分级
machine vision
Xinjiang Fuji apple
sugar degree
weight
grading
