SVM方法在淀粉分类问题中的应用

Application of SVM method in starch classification

采用不同品牌的马铃薯淀粉和玉米淀粉共计49个样品,运用VERTEX70进行光谱扫描,在不同光谱范围内,通过对原始光谱进行不同的预处理,得到淀粉样品的近红外光谱数据。在Matlab6.5仿真环境下,采用SVM工具包实现对样品数据的训练和预测,选取不同的核函数和惩罚因子C,可以准确地将淀粉进行分类。实验结果表明,利用近红外技术结合支持向量机对淀粉类别进行判别是可行的。

The near-infrared spectral（NIR）data of starch in the different spectral range which was processed from the original spectra based on the 49 different brands potato and com starch by using of VERTEX70 were obtained.The SVM tool kit of Matlab6.5 was used to finish the training of sample data and forecasts.By selecting a different kernel function and penalty factor C,starch can be accurately classified.The results showed that it was feasible that using near infrared technology combined with support vector machines can discriminate starch.