基于高光谱图像的即食海参新鲜度无损检测（英文）

发布时间：2018-08-18 19:32

【摘要】：新鲜度是即食海参加工品质调控和贮藏品质监控的关键指标。针对感官评定和现有理化检测无法满足即食海参产品大批量、标准化、工业化生产问题,提出了一种基于高光谱图像的即食海参新鲜度快速无损检测方法,通过图像主成分分析和波段比运算相结合,优选特征波长和图像;依据海参腐败机理,建立图像纹理特征与即食海参新鲜度等级间的关联模型,实现即食海参新鲜度无损、快速评价。首先针对高光谱图像巨大的数据量展开降维研究。根据即食海参体壁光谱吸收特性,以具有明显化学吸收特征的波长(474和985nm)为分界点,获得包括全检测波段(400~1 000nm)在内的六个待处理波段,通过分段图像主成分分析实现待测波段的优选,利用权重系数和波段比图像运算,最终将686和985nm波段比图像确定为特征图像。面向特征图像的感兴趣区域(ROI),构建灰度共生矩阵(gray-level co-occurrence matrix,GLCM)、灰度梯度共生矩阵(gray-gradient co-occurrence matrix,GGCM)、改进的局部二元模式纹理描述子(local binary pattern,LBP),分别提取纹理参数作为输入,以挥发性盐基氮(total volatile basic nitrogen,TVB-N)检测为标准,建立经粒子群优化的BP神经网络(back propagation,BP)即食海参新鲜度判别模型,新鲜度等级判别准确率分别为90%,95%和80%。结果表明,即食海参高光谱图像灰度梯度共生矩阵的纹理特征用于新鲜度判别效果较好。为即食海参新鲜度快速无损检测方法研究和仪器开发提供了理论基础和数据支持。
[Abstract]:Freshness is the key index of quality control and storage quality control of instant sea cucumber. Aiming at the problem that sensory evaluation and physical and chemical examination can not meet the problems of large quantity, standardization and industrial production of instant sea cucumber products, a fast nondestructive testing method for the freshness of instant sea cucumber based on hyperspectral image is proposed. By combining principal component analysis and band comparison, the characteristic wavelength and image were selected, and the correlation model between texture features and freshness grade of instant sea cucumber was established according to the mechanism of sea cucumber corruption. Rapid evaluation. Firstly, the dimensionality reduction research is carried out for the huge data volume of hyperspectral images. According to the spectral absorption characteristics of the instant sea cucumber's body wall, the wavelength (474 and 985nm), which has obvious chemical absorption characteristics, was used as the boundary point to obtain six unprocessed bands, including the whole detection band (400-1 000nm). Based on the principal component analysis (PCA) of segmented images, the optimal selection of the bands to be measured is realized. Finally, the 686 and 985nm band ratio images are determined as feature images by using the weight coefficient and the band ratio image operation. The region of interest (ROI),) for feature image is used to construct gray-level co-occurrence matrix (GLCM), grayscale gradient co-occurrence matrix (gray-gradient co-occurrence), and improved local binary pattern texture descriptor (local binary pattern LBP). The texture parameters are extracted as input, respectively. Based on the standard of (total volatile basic nitrogenin TVB-N detection, a back neural network (BP) model for distinguishing freshness of sea cucumber was established. The accuracy of freshness grade discrimination was 90% and 80%, respectively. The results show that the texture feature of grayscale gradient co-occurrence matrix of ready-to-eat sea cucumber hyperspectral image can be used to distinguish freshness. It provides a theoretical basis and data support for the research and development of fast nondestructive testing method for the freshness of instant sea cucumber.
【作者单位】：国家海洋食品工程技术研究中心大连工业大学;辽宁省海洋食品加工技术装备重点实验室大连工业大学;
【基金】：Marine public welfare industry research funding(201505029) Liaoning Provincial Natural Science Foundation(201602055) National Natural Science Foundation of China(31501561)
【分类号】：TP391.41;TS254.7

【相似文献】