基于高光谱图像和激光共聚焦显微镜技术快速测定猪肉嫩度  

Rapid Detection of Pork Tenderness Based on Hyperspectral Imaging and Confocal Laser Scanning Microscopy

作　　者：朱瑶迪 申婷婷[2] 赵改名[1] 邹小波[2] 李苗云[1] 石吉勇[2] 

机构地区：[1]河南农业大学食品科学技术学院,郑州450002 [2]江苏大学食品与生物工程学院,江苏镇江212013

出　　处：《中国食品学报》2017年第11期239-244,共6页Journal of Chinese Institute Of Food Science and Technology

基　　金：高新技术发展计划国家863项目(2011AA100807);全国优秀博士基金资助项目(200968);国家自然科学基金项目(61301239);新世纪优秀人才项目(NCET-11-00986);江苏省杰出青年基金项目(BK20130010)

摘　　要：以背最长肌和腰大肌为对象,利用高光谱图像(HSI)和激光共聚焦显微镜技术(CLSM)研究猪肉嫩度。利用CLSM结合荧光染色观察纤维密度和直径等组织学特性参数,判断不同部位猪肉嫩度。通过主成分分析(PCA)对其图像信息进行PCA,然后对光谱信息进行预处理。依据图像信息优选出3幅特征图像,提取对比度、相关性、角二阶矩和一致性等4个基于灰度共生矩阵的纹理特征变量,利用K-最邻近法(KNN)建立预测模型,为验证HSI的判别度。结果表明:KNN模型校正集和预测集的识别率分别为91.24%,83.57%,识别出腰大肌较背最长肌嫩。腰大肌的纤维直径和密度分别比背最长肌细且密,其观察结果与HSI的判别结果相一致。利用HSI快速预测猪肉嫩度具有可行性。The psoas major and the longest muscle of pork were as example in this study. The confocal laser scanning microscopy（CLSM）, which is an improvement over the traditional light microscopy in that it has the capability to scan product at different depths, reduce artifacts, and produce higher resolution images, was used to research the muscle fibers diameter and density to further judge the reliability of the hyperspectral results. A hyperspectral imaging system, which consists of both a digital camera and a spectrograph, can acquire images with both high spatial and spectral resolution contents. Therefore, HSI may capture both spatial and biochemical information simultaneously so that the texture characteristic of predicting the pork samples tenderness could be much greater. First, hyperspectral images of 80 pork samples were captured by HSI system. Dimension reduction was implemented on hyperspectral data by principal component analysis is（PCA） to select 3 characteristic images. Next, 4 characteristic variables were extracted by texture analysis based on gray level co-occurrence matrix（GLCM）, which is effective tool to calculate the texture characteristic of the samples. They are contrast, correlation, angular condmoment and homogeneity, respectively. Thus 12 characteristic variables in total for 3 characteristic images were extracted. PCA was conducted on 12 characteristic variables, and the sixth PCs variables were extracted as the input of the discrimination model. The detection model of pork tenderness was constructed by K-Nearest Neighbor（KNN）, according to the reference results of pork tenderness by Warner-Bratzler method. Detection results of KNN model are 91.24% and 83.57% in calibration and prediction sets, respectively. In addition, the result is consistent with the result of HSI using CLSM. This work shows that it is feasible to detect pork tenderness by HSI technique. Results are encouraging and show the promising potential of hyperspectral technology for detecting pork tenderness.

关 键 词：食品检测 猪肉嫩度 高光谱图像技术 纹理特征 激光共聚焦显微镜 

分 类 号：O657.3[理学—分析化学] TS251.51[理学—化学]