Study On Nondestructive Detection Of Pork Freshness Based On Single Technology And Multiple Information Fusion Technology

Rapid and accurate measurement of meat freshness is very significant in solving the food quality and safety for consumer sake. There are many indexs in assessing the meat freshness, Sensory evaluation and physical-chemical testing can not meet the requirements of rapid on-line detection, and single nondestructive testing technique is also difficult to achieve accurate comprehensive evaluation of meat freshness. In this work, the spoilage bacteria were isolated from the chilled metamorphic pork, and the fresh pork samples inoculated with the spoilage bacteria were stored to spoilage. Then, the pork freshness indicators were detected, the correlation among the pork freshness indicators was further analyzed during bacteria spoiling process. At the same time, the near infrared (NIR) spectroscopy, hyperspectral imaging, and NIR combined with computer vision (CV) and electronic nose (E-nose) were employed for non-destructive detecting the internal and external characteristic indexes of pork freshness, which achieves the comprehensive accurate evaluation of pork freshness. The main points are summarized as follows:1. Isolation and identification of the dominant spoilage bacteria. Microorganism is the main cause of rottenness of meat.5strains dominant spoilage bacteria were isolated from the chilled metamorphic pork, which were identified as Bacillus fusiformis J4, Acinetobacter guillouiae P3, Enterobacter cloacae P5, Pseudomonas koreensis PS1and Brochothrix thermosphacta S5by the morphology, physiology, biochemistry and16S rRNA molecular biology. Then, the yield factor YTVB-N/CFU of the total volatile basic nitrogen (TVB-N) was used as quantitative indicators to assess the spoilage capability of the different spoilage bacteria. The results showed that P. koreensis PS1has the strongest spoilage capability in chilled pork, B. fusiformis J4and B. thermosphacta S5are the stronger spoilage ability, and A. guillouiae P3and E. cloacae P5are the weakest among them. Then, to provide the modeling experimental samples, the fresh pork samples were inoculated with these spoilage bacteria for simulating the pork metamorphic process and for non-destructive detecting of pork freshness in the later experiment.2. Analysis on the correlation among the pork freshness indexes during bacterial spoiling process. Freshness is an important parameter for assessing the quality and safety of meat, which its evaluation have many indicators. In order to find the most indicator for reflecting freshness of meat, in the experiment, the fresh pork samples inoculated with P. koreensis PS1were stored in a refrigerator at4℃, the pork freshness indexes were detected by sensory evaluation, physicochemical methods and microbiological methods, and the correlation of these freshness indexes were analyzed during bacterial spoiling process. The results showed that the changes of TVB-N were significant, furthermore, it is significant difference (P<0.01) between TVB-N and total sugar, protein, sensory score, elasticity, resilience and adhesiveness, etc. And some of the appropriate non-destructive technology could be employed to assess pork freshness. 3. Rapid detection of pork freshness based on NIR during bacterial spoiling process. NIR can directly reflect the TVB-N conternt and can quickly detect the meat freshness. In this experiment, the pork samples of different freshness were used for study target. First, the different indexes have specific spectrum in the NIR region, and spectrum strength changes obviously with the pork spoilage grade. Then, synergy interval partial least squares (siPLS) was performed to select characteristic spectral variables of different indexes (TVB-N, total sugar, total fat, protein, and TVC) in pork based on NIR spectral data preprocessed by standard normal variate (SNV). The back-propagation neural network (BP-ANN) and siPLS were developed comparatively the quantitative models for predicting the different index. The experimental results showed that the optimum characteristic spectra related to each index can be selected by siPLS. The multi-index in pork can be determined simultaneously by BP-ANN or siPLS model during bacterial spoiling process. And the performance of BP-ANN model on the different indexes is better than siPLS model. The BP-ANN model have better prediction results of the indexes of TVB-N, total sugar, total lipid, and protein, which the determination coefficient (Rp2) are all more than0.850, except as TVC (Rp2=0.717). The results shows NIR can rapidly detect the meat freshness.4. Non-destructive detection of pork freshness based on hyperspectral imaging technique (HSI) during bacterial spoiling process. In the paper, TVB-N was still as the evaluation index of pork freshness, HSI is an emerging non-destructive technique, which can obtain more information for assessing meat freshness. In this experiment, the hyperspectral image data were collected from the pork samples of different freshness. First,177spectral characteristic variables from the hyperspectral data were selected by SNV preprocessing combined with siPLS. Secondly, the characteristic wavelength images from the hyperspectral data were extracted comparatively using principal component analysis (PCA) and the genetic algorithm-synergy interval partial least squares (GA-siPLS) and5characteristic wavelength images by GA-siPLS are more relevant to TVB-N content in pork than PCA. Next,6statistical parameter based on gray statistical moments were extracted from each characteristic wavelength image, amount to30feature variables form image information. Finally, PCA was implemented on177spectra variables,30image variables and207(177+30) variables based on data-fusion, and the top principal components were extracted for developing the TVB-N prediction model using BP-ANN, respectively. The experimental results showed that the model based on data-fusion is superior to others, which was achieved with RMSECV=1.28mg/100g and Rc2=0.922in the training set, RMSEP=1.60mg/100g and Rp2=0.900in the prediction set.5. Comprehensive evaluation of pork freshness based on multiple information fusion technology during bacterial spoiling process. Based on single technique for determinating the meat freshness, the study was further theorized that pork freshness during bacterial spoiling process was comprehensively evaluated by integrating NIR, CV and E-nose technology. In this experiment, the pork samples with different freshness were collected for data acquisition (such as chemical composition, color, texture, and odor, etc.) by three different techniques, respectively. Then, the individual characteristic variables from each sensor data were fused in feature level. Next, PC A was implemented on the individual characteristic variables and different data fusion from3different sensors data, and the top principal components were extracted for developing the TVB-N prediction model with BP-ANN, respectively. The experimental results showed that the model based on data fusion of three technologies (NIR, CV and E-nose) is superior to others, which was achieved with RMSECV=1.46mg/100g and Rc2=0.984in the training set, RMSEP=2.73mg/100g and Rp2=0.953in the prediction set. It is feasible to evaluate comprehensively the freshness during pork spoilage by integrating NIR, CV and E-nose technology through this experiment, and the accuracy and robustness of the model from three sensors information fusion were better than the model from single sensors or two sensors information. The study results offered a reference that multi-sensors information was applied to evaluate comprehensive pork quality and safety.This research offers a new idea to detect pork freshness based on muti-technique information fusion, and there is also of great significance in ensuring pork quality and safety, as well as safeguarding consumer interests.