基于气体传感阵列的水果成熟度智能分类研究OA

[Purposes]Existing commercial electronic noses exhibit an insufficient feature extraction ability for complex mixtures of volatile organic compounds(VOCs).Limitations are also encountered in specific scenarios,such as agricultural product quality inspection.To address these challenges,an intelligent classification method for fruit based on a metal oxide semiconductor(MOS)gas sensor array was proposed.[Methods]A twelve-dimensional sensing array,comprising MOS gas sensors and digital temperature and humidity sensors,was designed to detect changes in complex mixtures across different fruit stages sensitively.A dynamic gas path acquisition device was utilized to gather sensor array response voltage data from four fruit types(banana,mango,lychee,and kiwi)at various ripeness stages.A fruit ripeness dataset,incorporating gas response,temperature,and humidity features,was established.Data preprocessing was conducted using techniques such as sliding window mean filtering.A time-frequency fusion Transformer model was developed.Finally,fast Fourier transform and inverse fast Fourier transform and residual connections were integrated into the model to enable intelligent classification and evaluation of fruit ripeness.[Results]Multiple comparative experiments were conducted to validate the performance of the time-frequency fusion Transformer model using the test set.An accuracy of 89.58％,precision of 89.94％,recall of 89.58％,F1 score of 89.56％,and Matthews correlation coefficient(MCC)of 88.67％were achieved by the model.These metrics were 5.41,5.24,5.41,5.63,and 5.86 percentage points higher than those of the classic Transformer model,respectively.VOCs trends across fruit ripeness stages were effectively captured by the proposed MOS gas sensor array and acquisition device.Efficient extraction of time-frequency features was enabled by the time-frequency fusion Transformer model.[Conclusions]The intelligent classification method for fruit ripeness based on MOS gas sensor arrays,which achieves high sensitivity response to VOCs,utilizes artificial intelligence technology to realize non-destructive and rapid classification of perishable fruits,as well as their ripeness classification.This method has broad application prospects in areas such as intelligent warehousing and quality control.