目的建立基于可视传感器阵列的鳊鱼新鲜度评价模型。方法研究采用可视传感阵列与鱼体进行无接触式反应,提取阵列反应前后的颜色变化信息来表征鱼的气味特征;同时根据行业标准SC/T3032-2007测得表征鱼新鲜度的挥发性盐基氮(TVB-N)含量;将可视传感技术所得的特征信息与TVB-N指标含量进行关联,分别建立基于可视传感技术鱼新鲜度评价的定性模型BP神经网络和联合间隔偏最小二乘法(siPLS)定量模型。结果 BP神经网络模型精度较高,训练集正确率为86.79%,预测集正确率为86.43%;siPLS模型次之,模型校正集和预测集的正确率分别为82.52%和80.67%。结论可视传感器新技术所测得指标与TVB-N相关性较大,可快速预测出鱼在储藏期间TVB-N的变化从而能够快速、无损地评价鱼类新鲜度。 Objective To establish a freshness evaluation model based on visual sensor array. Methods The visual sensing array was used to make contactless reaction with fish body, and the color change information before and after the array reaction was extracted to characterize the odor of fish. At the same time, the volatility of fish freshness was measured according to the industry standard SC / T3032-2007 (TVB-N). The characteristic information of visual sensing technology was correlated with the content of TVB-N index to establish the qualitative model BP neural network and joint interval Partial least squares (siPLS) quantitative model. Results The precision of BP neural network model was higher, the correct rate of training set was 86.79%, and the correct rate of prediction set was 86.43%. The accuracy of siPLS model, model calibration set and prediction set were 82.52% and 80.67% respectively. Conclusion The results of the new visual sensor technology are highly correlated with TVB-N, which can quickly predict the change of TVB-N during storage so that the freshness of fish can be evaluated quickly and non-destructively.