基于光纤式高光谱成像技术的活体滩羊肉品质预测方法研究

Tan sheep is the superior and featured livestock in Ning xia. Its farming methods have changed from grazing to stabling since the implementation of grazing prohibition in Ning xia in 2003, which leads to an increase in production but a decrease in quality of single sheep. It shows that the fat content and smell of mutton have rised, while indexes such as lean, color, tenderness and juiciness have decreased. The original qualities of Tan sheep including delicacy, light taste and unique flavor have disappeared.Since quality of Tan sheep mainly depends on ante-mortem control, it is urgent to break through quality control technology of living Tan-sheep. In this project, living Tan-sheep’s subcutaneous hyperspectral images were collected with optical fiber as a medium ,the RGB images and spectra in specific areas were extracted. With the test data of colorimeter and the texture analyzer, spectral information corresponding to color, tenderness and juiciness were obtained by image process and spectral analysis. The key components to smell of mutton for 4-methyloctanoic acid and 4-methylnonanoic acid were measured by GC-MS. The measured values and spectral values were combined to get the corresponding spectral information of smell of mutton. Thus a prediction method based on feature-level information fusion method and fuzzy comprehensive evaluation for meat quality of living Tan-sheep was built.It intends to provide methodological guidance for the development of quality control technology of living Tan-sheep, which helps to improve meat quality in time by diet formulations adjustments and fat consumptions through movement.

滩羊为宁夏优势特色畜种。自2003年宁夏实行封山禁牧以来，滩羊养殖由放牧改为舍饲，单只羊产肉量上升，但品质下降，表现为脂肪含率、膻味上升，瘦肉率、色泽、嫩度、多汁性指标下降，使滩羊失去了肉质细嫩，膻味淡，风味独特的原有品质。优良的滩羊肉品质主要取决于宰前控制，因此，宁夏滩羊产业亟需突破活体羊肉品质调控技术。本项目拟以光纤为媒介，采集活体滩羊皮下高光谱图像，提取特定区域的RGB图像及光谱，通过图像处理及光谱分析，对照色差仪、质构仪检测数据，获得与活体滩羊肉色泽、嫩度、多汁性指标对应的光谱信息；将采用气-质联用对表征膻味的关键性成分4-甲基辛酸和4-甲基壬酸的实测值与光谱值结合，获得膻味的对应光谱信息。从而，建立基于特征级信息融合和模糊综合评判技术的活体滩羊肉品质预测方法。项目预期成果将为开发活体滩羊肉品质调控技术提供方法指导，使滩羊养殖户通过日粮配方调整、运动消耗脂肪等，及时改善肉品质。

（1）高光谱搭建.高光谱含成像光谱仪、光源、CCD相机和计算机。.（2）滩羊肉高光谱数据采集.G2D-COS、IVISSA得到MetMb、嫩度为6个和123个特征波长；IVISSA提取蛋白质和脂肪特征波长为46个和58个；SPA、CARS提取持水性、质构TBARS特征波段有5个和17个；G2D-CS-CARS提取TBA特征波长15个；IVISSA+VCPA选出IMP10个特征波长；IVISSA+VCPA、IVISSA+CARS选出HxR和Hx13和16个特征波长。.（3）滩羊肉品质评价.随贮藏时间延长，羊肉MetMb,TBARS增加；嫩度0.10-3.45，蛋白质0.17-0.26g/100g；脂肪0.01-0.11g/100g；硬度、黏性和咀嚼性呈上升趋势，精胺和亚精胺维持在3mg/kg和2mg/kg；色胺和β-苯乙胺为0.50-0.68mg/kg和0.47-0.53mg/kg；腐胺、尸胺和酪胺呈增加趋势。肌苷酸0.10-3.45umol/g，HxR0.07-3.69umol/g，Hx0.04-2.15umol/g。.（4）滩羊肉品质高光谱预测模型.G2D-COS+PLSR评估MetMb性能更好（R2C为0.842，R2P为0.849）；嫩度预测模型为OS-IVISSA-PLSR(Rc=0.85, Rp=0.79)；蛋白质和脂肪预测模型为1-SG-IVISSA-PLSR（Rc=0.91、Rp=0.82）, Normlize-IVISSA-PLSR（Rc=0.93、Rp=0.75）；咀嚼性预测模型为VCPA-LSSVM （Rp=0.814, RMSEP=352.85），硬度和粘聚性预测模型为IVISSV-VCPA-LSSVM(Rp=0.729, RMSEP=3606.807;Rp=0.856,RMSEP=938.134）；TBARS预测模型为OSC-CARS-LSSVM(Rc=0.91,Rp=0.91)；TBA、IMP、HxR、Hx预测模型为G2D-CS-CARS-LSSVM（Rc=0.92,Rp=0.91）、Detrending-CARS-MLR（Rc=0.93，Rp=0.96）；OSC-CARS-MLR（Rc=0.98，Rp=0.98）；De-trending-CARS-PLSR（Rc=0.96，Rp=0.96）。