基于多数据融合技术的腊肉品质分级方法

Quality Grading of Bacon Based on Multi-Data Fusion Technology

针对近年来备受关注的腊肉酸价和过氧化值超标、褪色、出油、发黏等品质问题,提出一种快速、准确、实用的检测技术。首先利用自组织特征映射网,对生化实验测得的酸价、过氧化值以及微生物菌落总数数据,在相关国家标准的基础上将样品腊肉的品质等级划分为4级:放心食用、可食用、不推荐食用和不可食用。在此基础上采用近红外光谱技术对腊肉的酸价与过氧化值进行检测,用遗传算法优选后的波长建模所得预测均方根误差分别是用优选前建模的41%、57%,所用波长数约为整个波段波长数的1/3。采用显微图像技术获得腊肉的菌斑信息,极大的改善了传统细菌总数检验方法操作复杂、主观性强、耗时长等问题。最后采用支持向量机对近红外光谱数据和显微图像数据进行多数据融合,建立腊肉可食用等级快速判别模型。

In recent years, problems with the quality of Chinese bacon such as acid value and peroxide value exceeding the standards, fading, oil exudation and sticky feeling to the touch have received growing attention. A fast, accurate and practical detection technology was developed to evaluate Chinese bacon quality. According to the relevant national standards as well as the results of acid value, peroxide value and total bacterial number in bacon samples measured by biochemical methods as the input of self-organizing feature map, the bacon samples were divided into four categories： safe to eat, edible, not recommended to eat and inedible. The acid value and peroxide value of bacon were detected using near infrared spectroscopy. The root mean square error prediction （RMSEP） results after the selection were 41% and 57% of those before selecting wavelengths by the genetic algorithms. The selected number of wavelength was 1/3 of the total number of the whole wavelength. Plaque area information was obtained by microscopic imaging technology, which has greatly improved many problems with traditional testing methods for the determination of total bacterial numbers, such as complex operation, subjectivity and time consuming. Finally, a quick discriminant model for grading the edibility of Chinese bacon was established using the support vector machine approach based on the near-infrared spectral data and microscopic image data.