摘要
为提高近红外光谱预测模型的精度和适用性,同时简化模型,提出了自适应蚁群优化偏最小二乘法优选特征波长的方法,建立不同产地苹果可溶性固形物含量混合分析模型.收集山东、陕西和新疆的富士苹果,采集3800 ～ 14000 cm-1范围的近红外光谱,并对其重要品质指标可溶性固形物含量进行测定.利用蚁群算法启发式全局搜索的特点,结合蒙特卡罗轮盘赌随机选择机制,优选苹果可溶性固形物含量的近红外光谱特征波长,然后用偏最小二乘法建立分析模型.与全光谱偏最小二乘模型和遗传偏最小二乘模型相比,蚁群优化算法选择的波长数最少,模型预测能力最强,预测的相关系数R和预测均方根误差RMSEP分别为0.9708和0.5144.研究结果表明,自适应蚁群优化算法可以有效选择近红外光谱特征波长,提高模型的稳健性和适用性.
Abstract Ant colony optimization algorithm combined with partial least squares （ACO-PLS） was employed to select the characteristic wavelength of near infrared （NIR） spectra for soluble solid content （ SSC ） in apple from different geographical region. The aim was to improve its accuracy and applicability, and to simplify the NIR prediction model. After collection of apple sample from three major apple-production regions of China, the original spectra of the apple in wavelength range of 3800 - 14000 cm 1 was acquired, and SSC was determined for reference measurements by standard method. Based on the features of heuristic global search and the random selection mechanism of Monte Carlo roulette, ACO explored optimally the efficient wavelength from the NIR spectroscopy of the apple to develop models for predicting the SSC of the apple. Experimental results showed that the performance of ACO-PLS model was superior to the performances from traditional PLS and GA-PLS models with the least variables. Good prediction performance was obtained for SSC with correlation coefficients of 0. 9708, and root mean square errors of prediction 0. 5144, respectively. The study demonstrates that adaptive ant colony optimization could effectively select the characteristic wavelengths of NIR spectral to improve the model robustness and applicability
出处
《分析化学》
SCIE
EI
CAS
CSCD
北大核心
2014年第4期513-518,共6页
Chinese Journal of Analytical Chemistry
基金
北京市自然科学基金重点项目(No.6144024)
公益性行业(农业)科研专项(No.201003008)
北京市农林科学院科技创新团队建设项目资助~~
关键词
近红外光谱
蚁群优化算法
特征波长
苹果
可溶性固形物含量
Keywords Near infrared spectroscopy
Ant colony optimization
Characteristic wavelength
Apple
Solublesolid content
