论文部分内容阅读
红、黄、紫马铃薯果肉颜色相近,肉眼很难识别,不同种类的马铃薯其营养成分各异,因此需要将马铃薯准确分类。传统的分类主要依靠化学实验方法,操作复杂且费时费力。利用高光谱技术对不同种类的马铃薯实施分类,不仅弥补传统方法的缺点而且能够快速、准确地实现分类。实验过程中首先利用高光谱成像系统采集黄、红、紫3种马铃薯的高光谱图像,并提取反射光谱数据;然后对光谱进行多元散射较正(MSC)预处理,接着运用主成分分析(PCA)选出7个特征波段;最后建立全波段、特征波段的支持向量机(SVM)和BP人工神经网络(BP-ANN)模型,实现马铃薯种类鉴别,准确率分别达到100%,说明利用高光谱图像技术能够准确的实现马铃薯分类。
The color of the red, yellow, and violet potato flesh is similar and the naked eye is hard to discern. The different types of potatoes have different nutrients and therefore require the accurate classification of the potatoes. The traditional classification mainly relies on chemical experiment methods, complicated operation and time-consuming and labor-intensive. The use of hyperspectral technology to classify different types of potatoes not only makes up for the shortcomings of traditional methods but also enables rapid and accurate classification. In the experiment, the hyperspectral images of yellow, red and purple potato were collected by hyperspectral imaging system and the reflectance spectra were extracted. Then, the spectra were pre-treated by multiple scattering and correction (PCA) Finally, the support vector machine (SVM) and BP artificial neural network (BP-ANN) model with full band and characteristic band were established to identify the type of potato with the accuracy of 100% respectively, indicating that the use of hyperspectral Image technology enables accurate potato classification.