追踪检测虾夷扇贝品质变化过程中的存活指标,生理指标以及电子鼻气味图谱的变化,建立保活流通过程中不同等级的活品虾夷扇贝电子鼻气味指纹图谱,购买市场上不同状态的活品虾夷扇贝,分别通过学习向量量化(learning vector quantization, LVQ)、概率(probabilistic neural networks, PNN)、支持向量机(support vector machine, SVM)神经网络对测试样品快速模式分类,最后通过对电子鼻传感器的筛选探索便携式快速品质鉴别设备的可能性。研究结果表明,24 h的极端胁迫环境放置较为完整的模拟了虾夷扇贝在保活流通过程中状态变差的过程；将电子鼻数据主成分分析、聚类分析结果与存活指标(开口率、缩边率以及死亡率)和生理指标(超氧化物歧化酶活性、耗氧率以及海水浊度)相结合可以把品质变化过程中的虾夷扇贝分成5个等级,并分别得到每个等级的扇贝气味指纹图谱；3种神经网络均可以对测试样品等级进行快速测定,其中支持向量机(SVM)神经网络兼具精确和快速的特点,测试样本T全部预测为等级4,测试样本N全部预测为等级3,从交叉验证到仿真预测所用时间仅为7.652 s；筛选得到的8个电子鼻传感器也可以对不同等级鲜活虾夷扇贝气味特征进行有效区分。

Optimization of live Yesso scallop transportation technology needs to establish an effective and special quality evaluation method, however, the complicated and variable physiological parameters of live creature make the quality difficult to identify.The objective of this study was to investigate whether an electronic nose, comprising 18 metal oxide Semiconductor gas sensors, could be used for measuring and modeling quality changes of Yesso scallop during live transportation, and to discuss the feasibility of portable rapid evaluation equipment for live Yesso scallop.Third instar live scallops, 8~12 cm, were purchased from a depuration workshop of Zhangzi Island, Dalian, China in May 2015.These scallops were just finished depuration and placed in stress environment(exposed in the air 24 h at 20 ℃, shocked and collided in a moving car), Samples were taken every 3 hours and recorded in chronological orderwere A, B, C, D, E, F, G, H, I and J, respectively.Two kinds of test scallops(T and N), 8~12 cm, arriving at port that day, were purchased from a fish market in Jungong Road of Shanghai, China in May 2015.Sample N were purchased at 4 am; Sample T, which had been placed in a foam box covered with crushed ice, were purchased at 2 pm.For each batch of scallops, 50 organisms were collected and three kinds of indicators were detected: survival indicators, physiological indicators and electronic nose odor fingerprints.Survival indicators included shell opening rate, skirt shrinking rate and death (skirt was unresponsive to stimuli) rate; Physiological indicators included SOD value, oxygen consumption rate and seawater turbidity; The Electronic nose utilized in this experiment was FOX 4 000 from Alpha MOS.Two grams of the scallop tissue sample (the samples were analyzed in septuplicate) were placed in a 10 mL volume of a vial and heated at 50 ℃ for 10 min, and 300 μl of headspace air was automatically injected into the e nose by a syringe and sensor responses were recorded for 120 s (flushing with reference air).The maximum response points of e nose, automatically recorded for each of 18 sensors, were used for analysis.The death rate, shell opening rate and skirt shrinking rate approached 58%, 88% and 100% after 24 hours of placing in stress environment.These scallops had completely lost commercial value, and 24 hours of placing in stress environment was supposed to simulate the whole process of Yesso scallop′s quality dropping.The result of principal component analysis and clustering analysis of electronic nose data was used to combine with survival indicators and physiological indicators, then the quality of scallop was divided into five grades, respectively were Degree 1(Sample A), Degree 2(Sample B), Degree 3 (Sample C and D), Degree 4(Sample E), Degree 5(Sample F, G, H and I.And standard electronic nose odor fingerprint was established for each grade.Two different grades of live scallop were tested by learning vector quantization neural network(LVQ), probabilistic neural network(PNN) and support vector machine neural network model(SVM) The training set were Sample A I, test set were Sample N and T These three kinds of neural network all can make a rapid evaluation of test samples, and the accuracy rates of LVQ network for training set was 100%, but accuracy rates for test set was not 100%, and the running time was 27.026 s.The accuracy rates PNN network for training set was 100%, but accuracy rates for test set was also not 100%, and the running time was 9.121 s.The result of SVM network was both accurate and rapid:test sample T were all forecasted to level 4, test sample N were all forecasted to level 3, and the time from cross validation to simulation was just 7.652 s.Eight electronic nose sensors that screened by loading value of principal component analysis can also be used to distinguish the live scallop odor characteristics.A rapid evaluation method for Yesso scallop during live transportation had been successfullyestablished in this study, and the screening of electronic nose sensors would be used toprovide technical support for developing the portable quality evaluation equipment in further research.

S24；S127

国家现代农业(贝类)产业技术体系建设专项(CARS-48-08B)