论文部分内容阅读
针对苹果近红外光谱数据的特点,研究了蚁群算法(ACO)在近红外光谱波长选择中的应用,建立了一种基于串联双通路构建图的波长变量选择模型。首先采集了苹果表面的漫反射近红外光谱,进而采用蚁群优化算法优选出近红外波长的最佳变量,使用所选择的近红外光谱波长数据建立苹果糖度预测模型。与GAPLS、siPLS等波长选择方法进行了比较,新模型的变量数减少到580,模型校正均方根误差RMSEC为0.2712,验证均方根误差RMSEP为0.3059。实验结果表明,蚁群算法用于苹果漫反射近红外光谱波长变量的选择,有效地减少了波长的使用,降低了模型复杂度,同时提高模型的预测精度。
According to the characteristics of near-infrared spectral data of apple, the application of ant colony algorithm (ACO) in wavelength selection of near-infrared spectrum was studied and a wavelength selection model based on tandem dual-channel construction graph was established. First, the diffuse reflectance near-infrared spectrum of apple surface was collected, and then the optimal variables of near infrared wavelength were optimized by using the ant colony optimization algorithm. The prediction model of apple sugar content was established by using the selected near infrared spectral wavelength data. Compared with the wavelength selection methods such as GAPLS and siPLS, the number of variables in the new model was reduced to 580, the root mean square error (RMSEC) was 0.2712, and the root mean square error of validation (RMSEP) was 0.3059. Experimental results show that the ant colony algorithm is suitable for the selection of wavelength of near-infrared spectrum of diffuse reflectance in apple, which can effectively reduce the use of wavelength, reduce the complexity of the model and improve the prediction accuracy of the model.