【摘要】 为了解决甜瓜含糖量预测中输入特征的"维数灾难"问题,提出了一种基于边际费希尔分析(MarginalFisher Model,MFA)和径向基神经网络(RBF)相结合的甜瓜含糖量无损检测方法.采用边际费希尔分析方法对高维的介电特性数据进行特征提取.该方法即能够保留高维空间中数椐的内蕴结构,又能同时进行降维,再应用RBF对低维特征数据进行训练和预测,结果表明该方法预测精度比其他算法有所提高.更多还原

【Abstract】 In order to solve the curse of dimensionality problem of the measurement of sugar content in cantaloupe,an algorithm based on Marginal Fisher Analysis(MFA) and Radial Basis Function(RBF) was proposed in this paper.The dielectric property data were projected from a high-dimensional space onto a low-dimensional space by MFA,this method can keep the intrinsic structure of the dielectric property data in high dimension space RBF is used to train and forecast the low-dimensional data.Experimental results show that the proposed algorithm performs much better than other traditional algorithms.更多还原