electronic nose characteristics of apples infected with different fungi

This dataset mainly contains data collected from apples infected with different fungi and fresh apples using an electronic nose, The electronic nose contains 8 sensors, and sensor No. 1 and sensor No. 5 use the same sensor to eliminate outliers in the data，If the difference between the 150-300s of the two sensors was greater than 1.2mg/L, the specimen will then be considered anomalous and removed, and the data for removing outlier samples using this method is stored in 'Data/sensors_eliminate' file. The "Fuji" apples selected in this dataset came from apple plantations in Gansu Province, China. 160 ripe apples were selected and randomly divided into 4 groups, 40 apples in each group, namely Group A, Group B, Group C and Group D; The fungi inoculated in the middle apples were Aspergillus niger, Penicillium expansum and Penicillium crustosum. The apple samples were pretreated with 75% alcohol on a sterile bench and dried at room temperature. Then, four holes were punched in four directions of each apple of the three groups A, B, and C with the inoculator. Sample apples were inoculated with 7-day-old molds through drilled loops, and the holes were covered with sterile film. The mold-inoculated apples were then placed in a 1000ml beaker, sealed with plastic wrap, and then placed in a 25°C constant temperature incubator for 5 days. Before the test, the apple samples were taken out of the incubator and placed for 30 minutes. To eliminate the influence of residual gas on the experimental results, electronic nose was cleaned with inert gas before using. Setting electronic nose parameters: cleaning time 500s, collection time 350s, sampling interval 1s, injection flow 150ml/min, the raw data store in 'Data/raw_data' file. Then Matlab is used to preprocess the raw data，The first is to smooth and filter the data, and use 3-point smoothing, 5-point smoothing, 7-point smoothing, 9-point smoothing and 11-point smoothing to smooth and filter the data after removing abnormal samples. The smoothed filtered data is stored in ‘Data/smoothed_data’ file. The second is feature extraction, we take the integral value, variance value, average differential value, maximum gradient value, relatively stable average value and energy value of the response curve of each sensor for 30-300s as the characteristic information of electronic nose. However, the value of 7NE/H2S-1000 and VOC-300 sensors is always 0 during the whole acquisition process, so this dataset only store the data measured by 6 sensors except 7NE/H2S-1000 and VOC-300, which stores in 'Data/feature_parameters_data' file, The second is to use Mahalanobis distance to propose abnormal samples in the data again, and the eliminated data is stored in ‘Data/Data/eliminate_anomalous_data‘ file. Finally, principal component analysis, factor analysis and linear discriminant analysis are used to reduce the dimension of the above data, and the data after dimension reduction is stored in ‘Data/dimensionality_reduction_data' file.

本数据集主要包含利用电子鼻采集的受不同真菌侵染的苹果与新鲜苹果数据。该电子鼻搭载8个传感器，其中1号与5号传感器为同款型号，用于剔除数据中的异常值。若两款传感器在150-300s区间内的检测差值大于1.2mg/L，则认定该样本为异常样本并予以剔除，采用该方法剔除异常样本后的数据存储于‘Data/sensors_eliminate’文件中。本数据集选取的富士（Fuji）苹果采自中国甘肃省苹果种植园，共挑选160个成熟苹果，随机分为A、B、C、D四组，每组40个；受试苹果所接种的真菌分别为黑曲霉（Aspergillus niger）、扩展青霉（Penicillium expansum）以及圆弧青霉（Penicillium crustosum）。所有苹果样本均在无菌工作台中用75%酒精进行表面消毒，并于室温下晾干。随后，针对A、B、C三组的每个苹果，使用接种针在其四向方位各打一个小孔，通过接种环将培养7天的霉菌接种至孔内，再用无菌薄膜封住孔洞。将接种霉菌的苹果放置于1000ml烧杯中，用保鲜膜密封后，移入25℃恒温培养箱中培养5天。实验开始前，将苹果样本从培养箱取出并静置30分钟。为消除残留气体对实验结果的干扰，使用惰性气体对电子鼻进行吹扫清洁。电子鼻参数设置如下：清洁时长500s，采集时长350s，采样间隔1s，进样流量150ml/min，原始数据存储于‘Data/raw_data’文件中。随后使用Matlab对原始数据进行预处理：首先对剔除异常样本后的数据进行平滑滤波处理，分别采用3点平滑、5点平滑、7点平滑、9点平滑以及11点平滑算法完成滤波，平滑滤波后的数据存储于‘Data/smoothed_data’文件中。其次开展特征提取工作，选取每个传感器在30-300s区间内响应曲线的积分值、方差值、平均微分值、最大梯度值、相对稳定平均值以及能量值，作为电子鼻的特征信息。但7NE/H2S-1000与VOC-300两款传感器在整个采集过程中数值始终为0，因此本数据集仅保留其余6款传感器的检测数据，该特征参数数据存储于‘Data/feature_parameters_data’文件中。随后采用马氏距离（Mahalanobis distance）再次对数据中的异常样本进行剔除，剔除后的样本数据存储于‘Data/Data/eliminate_anomalous_data’文件中。最后，采用主成分分析（Principal Component Analysis）、因子分析（Factor Analysis）以及线性判别分析（Linear Discriminant Analysis）对上述数据进行降维处理，降维后的数据存储于‘Data/dimensionality_reduction_data’文件中。