虾类产品新鲜度的高光谱图像特性及检测方法研究

The rapid, non-destructive and accurate determination of freshness of shrimp is critical for shrimp industry on dynamic assessment, grade marking and real-time monitoring of shrimp freshness during storage, distribution and marketing. Traditional detection methods of shrimp freshness are time-consuming and destructive, and cannot simultaneously obtain the information of physic-chemistry, microorganism indices, and sense organ indices relevant to freshness, resulting in difficultly detecting freshness of shrimp rapidly, non-destructively, and accuratly. In this project, hyperspetral images of shrimp in different freshness levels will be acquired by using a hyperspetral imaging system. On the basis of data mining and pattern recognition, the differences and variations of hyperspectral images will be explored for shrimps with different freshness, and the characteristics of hyperspectral images will also be revealed. The fingerprint image features will be extracted for expressing the information of chemical, microbiological, sensory indicators and establishing quantitative models for freshness prediction, resulting in realizing rapid, non-destructive, and accurate detection of frehssness of shrimps. The aim of this project is to establish a novel method from acquiring hyperspectral images, pattern recognition, simultaneously extracting fingerprint features relevant to physics and chemistry - microbial - sensory characteristics to rapidly, non-destructively, and accuratly determing freshness of shrimp. The output of this project will provide theoretical basis and methodological basis of designing hyperspectral imaging inspection equipments for detecting freshness of shrimps.

虾类新鲜度的快速无损准确检测，是实现贮藏、流通过程虾类新鲜度动态评估、等级标注和实时监控的基础和关键。常规检测方法测定速度慢，需破坏样品，且不能同步获取与新鲜度相关的理化、微生物和感官指标信息，无法实现虾类新鲜度的快速无损准确检测。本课题通过高光谱成像获取虾体高光谱图像信息，结合数据挖掘、模式识别等方法，探究不同新鲜度下虾类高光谱图像信号的差异及变化规律，揭示虾类新鲜度的高光谱图像特性，并准确提取能表达理化、微生物、感官指标信息的高光谱-图像特征指纹参数，构建虾类新鲜度预测模型，从而实现虾类新鲜度的快速无损准确检测。课题旨在建立一种从虾类高光谱图像信息获取、模式识别、理化-微生物-感官特征指纹参数同步提取，到虾类新鲜度快速无损准确检测的新方法，为研制虾类新鲜度高光谱成像检测仪器设备提供理论依据和方法基础。

虾类新鲜度的快速无损准确检测，是实现贮藏、流通过程虾类新鲜度动态评估、等级标注和实时监控的基础和关键。课题围绕虾类新鲜度指标的光谱特性与识别方法、虾类高光谱图像特征参数优化提取方法和高光谱图像预测建模方法三个方面开展了研究。（1）开展了基于南美白对虾挥发性盐基氮的鲜度近红外光谱识别方法研究，结果表明，利用光谱预处理、光谱降维手段获取光谱特征参数，结合SR、LSSVM机器学习分类方法，能较准确实现南美白对虾肌肉挥发性盐基氮含量（新鲜度等级）识别。（2）开展了虾类高光谱图像参数优化提取方法研究，提出了一种基于UVE-SPA 方法的虾类高光谱图像特征波段选择方法，结合该方法建立预测模型，对虾体内明胶含量的预测相关性系数可达0.965。此外，提出了一种基于自编码网络流形学习的光谱数据降维方法，该方法能有效提高虾类新鲜度等级光谱预测建模的精度，为虾类鲜度指标无损检测研究中高光谱特征波段选择提供了有效的手段。（3）开展高光谱图像预测建模方法研究，提出了一种基于稀疏表示的高光谱图像预测建模方法，为虾类高光谱成像的快速无损检测预测建模研究提供了一种全新的途径。本课题研究建立了从新鲜度高光谱图像特征参数提取，到虾类新鲜度快速无损准确检测预测建模的相关有效方法，为研制虾类新鲜度高光谱成像检测仪器设备提供理论依据和方法基础。