猪肉新鲜度的激光散斑图像检测方法研究

发布时间：2018-03-02 04:25

本文关键词： 猪肉新鲜度激光散斑散斑活性惯性矩互相关系数 LDA线性判别分析　出处：《江苏大学》2017年硕士论文　论文类型：学位论文

【摘要】：冷鲜猪肉是人们日常生活主要的肉类消费品种。由于食品安全事件频频发生,人们对猪肉品质日趋重视。传统的猪肉新鲜度检测方法主要为感官评定和实验室理化检测,前者检测结果不稳定,而后者检测过程耗时费力,且不适应现场快速检测,难以满足消费者的需求。而激光散斑技术具有快速、无损、简便、成本低等优势。本研究将激光散斑技术应用于冷鲜猪肉新鲜度的检测,具体研究内容如下:1、基于国内外学者对激光散斑技术的研究结果,确定了冷鲜猪肉新鲜度激光散斑检测的整体试验方案,设计并搭建了激光散斑试验装置,编写了基于VS2013+MFC开发的散斑图像批量采集及处理程序,并通过预实验确定了最佳试验参数,其中激光波长为465 nm和660 nm、激光功率为15.7 mW、采集时间为60 ms、激光入射角为30°。2、针对散斑图像处理过程的研究发现,散斑图像不同行的选取会对IM值产生影响,且传统IM算法受异常值干扰较大,为此提出三点改进:设计了排序算法,动态选择散斑活性最高峰及周围2个相邻行,依此计算样本IM值;改进共生矩阵的修正矩阵计算方法;改进非零元素偏离对角线距离的计算方法。结果显示,排序算法能快速定位IM最高峰位置,且改进后的方法可以有效抑制异常值干扰,更能真实反映出样本的活性差异。3、分别对冷藏期间猪肉失水量、肉色(L*、a*、b*)及TVB-N含量变化与惯性矩IM和互相关系数Ckτ值进行分析。发现猪肉失水量变化较小(0.14g/d)时,对样本散斑活性变化影响便较小,而失水量变化较多(2g/d)时,水分散失导致样本散斑活性明显降低,这说明水分是影响散斑活性变化的主要原因。猪肉表面颜色与散斑活性指标均呈正相关关系,其中a*值与散斑活性指标间的相关性最大(0.8),说明散斑活性能反映出猪肉的新鲜度变化。猪肉冷藏期间的TVB-N含量与散斑活性指标呈负相关关系,因此仅依靠散斑活性无法对TVB-N含量做出准确预测。4、选取465 nm和660 nm两种波长,基于IM和Ckτ值两种散斑活性指标,建立猪肉新鲜度LDA线性判别模型。结果显示,465 nm波长下两种指标的单主成分判别模型,其识别率都要高于660 nm,说明465 nm波长更能反映猪肉新鲜度变化。当选择两种波长下的IM值和465 nm波长下第21帧及第201帧散斑图像的Ckτ值四个特征参数作为主成分建模时,模型的识别率最好。其训练集和预测集能分别达到95.31%和96.88%,且能完全识别腐败肉样本,因此利用激光散斑技术检测冷鲜猪肉新鲜度的方法具有可行性。
[Abstract]:Cold fresh pork is the main meat consumption variety in people's daily life. People pay more and more attention to pork quality because of the frequent food safety incidents. The traditional methods of pork freshness detection are mainly sensory assessment and laboratory physical and chemical testing. The former is unstable, while the latter is time-consuming and difficult to meet the needs of consumers because it is not suitable for quick detection on the spot. The laser speckle technique is fast, non-destructive and simple. In this study, the laser speckle technique was applied to the detection of freshness of chilled pork. The specific research contents are as follows: 1. Based on the research results of laser speckle technology by domestic and foreign scholars, The whole test scheme of laser speckle detection for freshness of chilled pork was determined, the laser speckle test device was designed and built, and the batch acquisition and processing program of speckle image based on VS2013 MFC was compiled. The optimum experimental parameters were determined by pre-experiment. The laser wavelength was 465 nm and 660 nm, the laser power was 15.7 MW, the acquisition time was 60 Ms, and the incident angle was 30 掳. The selection of different speckle images will affect the IM value, and the traditional IM algorithm is greatly disturbed by the outliers. For this reason, three improvements are put forward: a sorting algorithm is designed to dynamically select the highest peak of speckle activity and two adjacent rows around it. Based on this, the IM value of the sample is calculated; the modified matrix calculation method of symbiosis matrix is improved; and the calculation method of non-zero element deviation from diagonal distance is improved. The results show that the sorting algorithm can locate the peak position of IM quickly. The improved method can effectively suppress the abnormal value interference, and can reflect the activity difference of the sample. When the change of water loss of pork was smaller than 0.14g / d, the effect on the speckle activity of the sample was smaller, but the change of water loss was more than 2 g / d, when the change of TVB-N content, the moment of inertia (IM) and the correlation number of C _ k 蟿 were analyzed, the results showed that, when the change of pork water loss was less than 0.14g / d, the change of sample speckle activity was less, but the change of water loss was more than 2 g / d. The loss of moisture resulted in a significant decrease in speckle activity, which indicated that moisture was the main reason for the change of speckle activity, and there was a positive correlation between the color of pork surface and the index of speckle activity. The correlation between a * value and speckle activity index was the most significant, which indicated that the speckle activity could reflect the freshness of pork. The TVB-N content of pork was negatively correlated with the speckle activity index during cold storage. Therefore, speckle activity alone can not accurately predict the content of TVB-N. The wavelength of 465nm and 660nm is selected, based on two speckle activity indexes of IM and Ck 蟿, The LDA linear discriminant model of pork freshness was established. The results showed that the single principal component discriminant model of two indexes was established at 465nm. The recognition rates are higher than 660 nm, indicating that 465nm wavelength can better reflect the variation of pork freshness. When selecting the IM value of two wavelengths and the four characteristic parameters of frame 21 and frame 201 speckle image at 465nm as principal component modeling, The model has the best recognition rate, its training set and prediction set can reach 95.31% and 96.88, respectively, and it can fully identify the rotten meat samples. Therefore, the method of detecting the freshness of cold fresh pork by laser speckle technique is feasible.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O657.3;TS251.51

【参考文献】