基于随机森林分类模型的葡萄干特征分析

Characterization of Raisins Based on Random Forest Classification Model

为了实现两种葡萄干的高效率分类,以R语言作为工具,将两种土耳其葡萄干(Besni和Kecimen)的900颗(每种450颗)葡萄干图像数据作为数据集,通过图像提取技术,提取7种形态学特征:Area、Perimeter、MajorAxisLength、MinorAxisLength、Eccentricity、ConvexArea、Extent,数据集经过归一化和清除噪音的处理,选择随机森林算法建立分类模型,与SVM模型相比较,结果表明:随机森林模型使用混淆矩阵进行综合评价结果显示与SVM模型不分上下,但对于葡萄干数据而言,使用随机森林模型对变量重要性的解读更适合,研究表示Perimeter和MajorAxisLength这两个形态学特征对随机森林的分类模型十分重要。

In order to realize the efficient classification of two kinds of raisins, R language was used as a tool, and the image data of 900 raisins (450 raisins each) of two kinds of Turkish raisins (Besni and Kecimen) were used as a dataset, and seven morphological features were extracted by image ex-traction technique: Area, Perimeter, MajorAxisLength, MinorAxisLength, Eccentricity, ConvexArea, and Extent, the dataset was normalized and noise removal, and the Random Forest algorithm was selected to build the classification model, which was compared with the SVM model, and the results showed that: the Random Forest model using the confusion matrix for the comprehensive evalua-tion of the results showed that it was indistinguishable from the SVM model, but for the raisin data, the interpretation of the importance of the variables using the random forest model is more appro-priate, and the study indicated that the two morphological features of Perimeter and MajorAx-isLength are important for the classification model of the random forest.