基于多尺度显微高光谱成像技术的猪肉新鲜度无损检测研究

本研究把高光谱成像技术拓展到微观尺度领域，综合高光谱成像技术和显微成像技术的优势，从宏观、微观两个层面对猪肉内外特征进行可视化分析，在特定尺度下和特定波段范围内来跟踪猪肉的腐败过程，解析猪肉在腐败变质过程中其内部组织结构、致腐微生物数量和分布的变化规律。实验以猪的里脊肉为试验样本，首先通过感官评定和理化分析综合确定样本的新鲜度；然后，构建一个显微高光谱成像系统并在不同显微尺度下获取样本高光谱图像，优化数据挖掘算法，对原始数据进行滤噪、降维，从海量数据中提取图像特征和光谱特征，通过不同尺度的微观图像揭示猪肉的新鲜度与猪肉组织结构的相关性；通过不同波长图像揭示猪肉的新鲜度与猪肉光学特性的相关性。最后利用多信息融合方法和模式识别方法构建猪肉新鲜度评价模型。本项目的研究将为猪肉及其制品在运输、贮藏及加工过程质量与安全的实时控制提供理论依据；对保障肉产品质量与安全，维护消费者利益有着重要的现实意义。

The hyperspectral imaging technique is extended to the micro scale field in this study, and the external and internal attributes of pork meat can be visualized from macros/ micro level, using integrating hyperspectral imaging and micro-imaging. Micro-hyperspectral imaging was proposed to explore the mechanism of pork meat spoilage using visual analysis of the change of tissue structure and spoilage bacterial distribution in the pork. Herein, pig tenderloin was studied in the experiment, and first its freshness was determined, as reference results, using human panel test and chemical analysis. Next, we constructed a micro-hyperspectral imaging system to acquire hypercube in different micro-scale, and then a series of data processing algorithms including noise filtering, dimension reduction and feature extraction were implemented on raw data. Thus, the correlation between pork meat freshness and pork tissue was revealed from the multi-scale micro-images; the correlation between pork meat freshness and its optical properties was revealed from the hyperspectral images at the specific bands. Finally, we will construct a mathematical model for pork freshness evaluation using data-fusion and pattern recognition algorithm. This research achievement will provide a theoretical basis for the real-time control in pork products processing, storage and transportation, and it is of great significance to ensure the quality and safety of meat products and safeguard the interests of consumers.