基于LS-SVM高光谱成像鱼新鲜度鉴别  

Classification of Fish Freshness Based on LS-SVM and Hyperspectra Imaging Technology

作　　者：章海亮[1] 叶青[1] 罗微[1] 刘雪梅[1] Zhang Hailiang Ye Qing Luo Wei Liu Xuemei(School of Electrical and Automation Engineering, East China Jiaotong University, Nanchang 330013, China)

机构地区：[1]华东交通大学电气与自动化工程学院,江西南昌330013

出　　处：《华东交通大学学报》2017年第2期105-111,共7页Journal of East China Jiaotong University

基　　金：国家自然科学基金项目(61565005);江西省科技支撑项目(20142BDH80021)

摘　　要：基于高光谱成像光谱信息的鱼新鲜度(鱼不同冷冻时间以及冻融次数)鉴别。首先,提取鱼样本感兴趣区域(region of interest,ROI)光谱,分别采用蒙特卡罗无信息变量消除(Monte Carlo free information variable elimination,MCVE),连续投影算法(successive projections algorithm,SPA)和随机青蛙算法(random frog,RF)提取特征波长,三种算法分别得到90,31和49个特征变量,采用最小二乘支持向量机作为分类模型,将90,31和49个特征变量作为LS-SVM模型的输入变量建立分类模型,基于SPA-LS-SVM和MCVE-LS-SVM模型预测集识别率都达到了98%,而采用RF-LS-SVM建立的模型取得了较差的预测结果 ,模型预测集识别率都只是达到了88%。结果表明,SPA-LS-SVM作为分类模型优于其他模型,SPA选择的特征波长,不但可以简化模型,还可以提高模型的预测精度,基于高光谱成像技术可以用于鱼新鲜度(鱼不同冷冻时间以及冻融次数)鉴别。This study investigated the feasibility of using near infrared hyperspectral imaging system（NIR-HIS）technique for non-destructive identification of fresh and frozen-thawed fish fillets. Hyperspectral images of freshness, storage time, and frozen-thawed times of fillets for turbot flesh were obtained in the spectral region of381~1 023 nm. Reflectance values were extracted from each region of interest（ROI） of each sample. Monte Carlo free information variable elimination（MCVE） algorithm, successive projections algorithm（SPA） and random frog（RF） were carried out to identify the most significant wavelengths. Based on the ninety, thirty-one and fortynine wavelengths suggested by MCVE, SPA and RF, respectively, two classified models（least squares-support vector machine, LS-SVM and SIMCA） were established. Among the established models, SPA-LS-SVM model performed well with the highest classification rate（100%） in calibration and 98% in prediction sets. SPA-LSSVM and MCVE-LS-SVM models obtained better results 98% of classification rate in prediction set with thirtyone and ninety effective wavelengths respectively. The RF-LS-SVM model obtained poor results with 88% of classification rate in prediction set. The results showed that NIR-HIS technique can be used to identify the varieties of fresh and frozen-thawed fish fillets rapidly and non-destructively, and SPA was effective wavelengths selection method.

关 键 词：蒙特卡罗无信息变量消除 连续投影 随机青蛙 LS-SVM 

分 类 号：O433[机械工程—光学工程]