用遗传算法提取南疆红枣总糖近红外光谱特征波长剖析.docVIP

用遗传算法提取南疆红枣总糖的近红外光谱特征波长 彭云发1，詹映1，彭海根1，刘飞1，罗华平1,2* (1.塔里木大学机械电气化工程学院， 新疆 阿拉尔 843300；2.新疆维吾尔自治区普通高等学校现代农业工程重点实验室， 新疆 阿拉尔 843300) 摘 要：本研究尝试利用近红外光谱技术测量红枣的总糖含量，针对采用偏最小二乘（PLS）法建立近红外光谱预测模型时波长筛选问题，提出用联合区间偏最小二乘法（siPLS）与遗传算法（GA）相结合的方法遗传联合区间偏最小二乘法（GA-siPLS）来提取近红外光谱特征区域和特征波长，提高模型预测精度的方法。结果表明：将全谱等分成20个子区间，用联合区间偏最小二乘法优选出4个特征子区间，在这4个子区间的基础上再用遗传偏最小二乘法继续筛选出12个特征波长。用12个特征波长建立的偏最小二成模型精度要好于全谱建立的模型，其主因子数减少了4个，预测集标准偏差（RMSECP）减少了25%，预测相关系数（RP）提高了5%。该方法选取的波长变量建立的校正模型，不仅使模型简洁、优化，而且增强了模型的预测能力。 关键词：近红外光谱,特征波长,联合区间偏最小二乘法,遗传算法,红枣 Analysis of near infrared spectroscopy of jujube sugar content by genetic algorithms PENG Yun-fa1, ZHAN Ying1, PENG Hai-gen1, LIU Fei1, LUO Hua-ping1,2* (1 School of Mechanical and Electrical Engineering, Tarim University, Alar Xinjiang 843300; 2 The Key Laboratory of Colleges Universities under the Department of Education of Xinjiang Uygur Autonomous Region, Alar Xinjiang 843300) Abstract: This study was conducted to attempt to measure the total sugar content of jujube (Ziziphus jujuba cv. Huizao) using near-infrared spectroscopy, the information can be abstracted by partial least-square regression (PLS). In order to select wavelengths of near infrared spectroscopy in the prediction model of partial least squares regression of jujube sugar content detection, correlation coefficients and genetic algorithms and interval partial least squares (GA-siPLS) were used to select wavelength from near infrared spectroscopy in partial least squares regression model. The spectra were divided into twenty intervals, among which four subsets were selected by GA-siPLS to characterize the net signals of jujubes’ saccharinity. Then twelve absorbance values(A) at their characteristic wavelengths were screened out. It showed that the model developed by selecting twelve wavelengths was better than that of full spectrum．the principal factor number reduces by 4 and the root mean square error of prediction of prediction set reduces by 25%, while correlation coefficient of prediction impro