Research On Prediction Technology Of Apple Freshness Based On Sensor Array

Apple is the favorite of consumers in different countries because of its taste and high nutritional value.The output and exports of apple rank among the top in the world.However,due to improper storage and other factors,apple picking to sales in the whole process,its appearance,texture,taste will change,wilting,rot and a series of problems cannot be avoided,which to fruit farmers and even the country caused economic losses,but also to consumers harm.If the process of apple from fresh to rot can be predicted,the storage risk of apple can be reduced effectively and the economic loss caused by apple rot can be avoided.Aiming at the problems of low accuracy of existing prediction technology and damage to samples caused by detection process,this study takes the prediction of apple odor characteristics as the starting point,and finally realizes the prediction of apple freshness with high accuracy and non-destructive.The odor information of ethylene,carbon dioxide and ethanol emitted during apple storage is an important characteristic of apple freshness.An electronic nose based on gas sensor array was designed in order to accurately predict the freshness grade of apples.According to the physiological effects of apples after picking and previous studies of the research group,sensors corresponding to gases closely related to apple freshness were selected to design sensor arrays,and Zig Bee technology was used as the data transmission method to complete the odor acquisition system.Apples with different degrees of freshness have different odor characteristics,which can be predicted by accurate prediction and recognition of odor characteristics.Therefore,a back-propagation(BP)neural network prediction model based on the modified Sparrow search algorithm(SSA)based on chaotic sequence is proposed.The change of apple odor information over time was studied to complete the prediction of apple odor characteristics.By fitting the relationship between the prediction determination coefficient and the input vector,the accuracy benchmark of the prediction model was set,which further improved the prediction accuracy of apple odor information.Finally,the recursive neural network(ELMAN)suitable for classification prediction was selected to predict the freshness grade of apples by using their odor characteristics,and the final prediction result of apple freshness was obtained.The variation of the characteristic value of the freshness in the first five days of the storage period was used to predict the freshness in the next 30 days.The experimental results show that the prediction determination coefficient of the optimized BP neural network model is 0.95851,which is significantly improved from 0.8 before optimization.The ELMAN freshness grade prediction model predicted 100% of the apple freshness on day 6 and 80% of the apple freshness on day 30.The prediction model has the characteristics of high accuracy and reliable results.At the same time,the damage to the apple is avoided in the process of data acquisition,and the non-destructive testing is realized.