基于Fisher判别分析对苹果新鲜度的识别研究

以淘宝销量前五的苹果（天水花牛、阿克苏糖心、洛川红富士、奶油富士、黄元帅）为研究对象，置于20 ℃和55%相对湿度条件下贮藏，在贮藏期的第0、5、10、15、20、25 d分别对其含水量、硬度、色差、可溶性固体含量（SSC）及可滴定酸进行测定，并对五种苹果进行感官评定、聚类分析及Fisher判别。结果表明，感官评定及聚类分析将苹果新鲜度分为新鲜、次新鲜、可接受和腐败。通过Fisher判别分析法得出苹果在贮藏期的2个判别函数，函数1的总变异系数为93.90%（P=0.000），函数2的总变异系数为6.10%（P=0.003），两个函数均有判别意义（P<0.05）。将样品分为测试集和训练集，训练集的识别率为96.25%，测试集的识别率为92.5%。综上，以本文所选取的理化指标作为判别因子所建立的Fisher判别分析模型可以有效识别苹果的新鲜度。

Recognition of Apple Freshness Based on Fisher Discriminant Analysis

Taking the top five apples sold on Taobao (Tianshui Huaniu, Aksu Tangxin, Luochuan Red Fuji, Cream Fuji and Huang Marshal) as the test objects, store at 20 ℃ and 55% relative humidity, on the 0th, 5th, 10th, 15th, 20th and 25th days of the storage period to determine its water content, hardness, color difference, soluble solid content (SSC) and titratable acid, and the sensory evaluation, cluster analysis and Fisher discrimination were carried out for five kinds of apples. Result showed that, the freshness of apples was divided into fresh, sub-fresh, acceptable and spoiled by sensory evaluation and cluster analysis. Two discriminant functions of apples during storage were obtained by Fisher discriminant analysis. The total coefficient of variation of function 1 was 93.90% (P=0.000), the overall coefficient of variation of function 2 was 6.10% (P=0.003), and both functions had discriminant significance (P<0.05). The samples were divided into test set and training set, the recognition rate of training set was 96.25%, and the recognition rate of test set was 92.5%. The results showed that the Fisher discriminant analysis model established with the physical and chemical indicators selected in this paper as the discriminant factors could effectively identify the freshness of apples.