基于高光谱技术的籼稻霉变程度鉴别模型构建与优化
发布时间：2018-08-09 15:04
【摘要】：为解决快速无损鉴别籼稻霉变程度问题,利用高光谱技术采集200份霉变样本可见/近红外光谱信息,随机选取155份样本作为校正集,剩余45份作为验证集,根据预测浓度残差检验标准对校正集中异常样本进行剔除。以新校正集建立主成分线性判别分析(PCA-LDA)和簇类独立软模式法(SIMCA)模型,选用正确识别率为指标,优选最佳鉴别模型。并采用连续投影算法(SPA)提取特征波长,优化优选的最佳模型构建速度。研究结果表明,PCA-LDA对所有样本的误判总数为15,正确识别率为92.50%;SIMCA和SPASIMCA对所有样本的未能正确识别总数分别为6、2,正确识别率分别为97.00%、99.00%,并且经SPA筛选的变量数为20,仅占原始变量数的7.81%,建模时长缩短为原始变量的40.93%。因此,SPA-SIMCA鉴别效果最好,该方法在快速、准确鉴别籼稻霉变程度上具有可行性。
[Abstract]:In order to solve the problem of rapid nondestructive identification of mildew degree of indica rice, 200 musty samples were collected by hyperspectral technique. 155 samples were randomly selected as calibration set, and the remaining 45 samples were used as validation set. According to the test standard of predicted concentration residual, the abnormal samples in correction concentration are eliminated. The principal component linear discriminant analysis (PCA-LDA) and cluster independent soft mode method (SIMCA) model are established by using the new correction set. The best discriminant model is selected by selecting the correct recognition rate as the index. The continuous projection algorithm (SPA) is used to extract the feature wavelength to optimize the optimal model construction speed. The results show that the total number of misjudgments for all samples by PCA-LDA is 15, the correct recognition rate is 92.50% for Simca and SPASIMCA for all samples is 6 / 2, respectively, and the correct recognition rate is 97.00 / 99.00, and the number of variables screened by SPA is 20, accounting for only the original number. The initial variable number is 7.81, and the modeling time is reduced to 40.933 of the original variable. Therefore, SPA-SIMCA is the best method for identification of mildew of indica rice.
【作者单位】： 中南林业科技大学机电工程学院;华南农业大学南方农业机械与装备关键技术教育部重点实验室;
【基金】：国家自然科学基金(31401281) 湖南省科技计划重点研发项目(2016NK2151) 湖南省自然科学基金(14JJ3115) 湖南省高校科技创新团队支持计划(2014207)
【分类号】：O657.3;TS210.7
本文编号：2174469
[Abstract]:In order to solve the problem of rapid nondestructive identification of mildew degree of indica rice, 200 musty samples were collected by hyperspectral technique. 155 samples were randomly selected as calibration set, and the remaining 45 samples were used as validation set. According to the test standard of predicted concentration residual, the abnormal samples in correction concentration are eliminated. The principal component linear discriminant analysis (PCA-LDA) and cluster independent soft mode method (SIMCA) model are established by using the new correction set. The best discriminant model is selected by selecting the correct recognition rate as the index. The continuous projection algorithm (SPA) is used to extract the feature wavelength to optimize the optimal model construction speed. The results show that the total number of misjudgments for all samples by PCA-LDA is 15, the correct recognition rate is 92.50% for Simca and SPASIMCA for all samples is 6 / 2, respectively, and the correct recognition rate is 97.00 / 99.00, and the number of variables screened by SPA is 20, accounting for only the original number. The initial variable number is 7.81, and the modeling time is reduced to 40.933 of the original variable. Therefore, SPA-SIMCA is the best method for identification of mildew of indica rice.
【作者单位】： 中南林业科技大学机电工程学院;华南农业大学南方农业机械与装备关键技术教育部重点实验室;
【基金】：国家自然科学基金(31401281) 湖南省科技计划重点研发项目(2016NK2151) 湖南省自然科学基金(14JJ3115) 湖南省高校科技创新团队支持计划(2014207)
【分类号】：O657.3;TS210.7
【相似文献】
相关硕士学位论文 前1条
1 金灵;基于高光谱技术的宁夏滩羊肉内部品质检测方法研究[D];宁夏大学;2014年
，本文编号：2174469
本文链接：https://www.wllwen.com/kejilunwen/huaxue/2174469.html
教材专著
