南丰蜜桔可溶性固形物非线性模型研究

探索应用近红外透射光谱无损检测技术，结合径向基神经网络和反向传播神经网络两种人工智能算法，建立南丰蜜桔可溶性固形物快速无损检测的非线性稳定预测模型。通过43个未参与建模的南丰蜜桔样品作为外部验证数据，验证径向基函数神经网络模型的性能，预测模型的相关系数R=0．92。研究结果表明：径向基函数神经网络比反向神经网络获得的预测结果更好，校正模型的预测均方根误差从0．72^0Brix降低到0．65^0Brix。径向基函数神经网络方法是解决近红外光谱检测南丰蜜桔可溶性固形物中非线性建模问题的有效方法。

A Research on Nonlinear Model of Soluble Solids Content Based on Radial Basis Function Neural Network in Nanfeng Mandarin Fruits

The research aims to develop nonlinear mathematical model for rapid determining soluble solids content （SSC） of Nanfeng mandarin fruit nondestructively using near infrared （NIR） transmittance spectroscopy technology, combining with artificial intelligent algorithms of basis function neural network （RBFNN） and back propagation neural network （BPNN）. The 43 unknown Nanfeng mandarin fruits are used for external verifying data, the performance of RBFNN model is verified with Correlation coefficient （R = 0.92）. The experimental results show that the predicting result of RBFNN is better than that of BPNN. Root mean square error of prediction （RMSEP） decreases from 0.72 ^0Brix to 0.65 ^0Brix. So RBFNN method is an effective method to solve the nonlinear problem in NIR spectra measurement SSC of Nanfeng mandarin.