Research On Apple Sweetness Recognition Technology Based On Gas Sensor Array

China's apple production ranks among the highest in the world,the annual production of apple can reach more than 40 million tons,but the annual export volume is only more than 1 million tons,accounting for only 2.5% of the total,the main reason is the low degree of apple quality classification.In the actual sales process of apples,apples with bright colors,clear textures and uniform sizes are manually screened out for sale as high-quality fruits.However,due to the influence of experience,emotions and other factors,manual screening can't accurately judge the actual sweetness of apples,which affects sales and exports.Therefore,it is necessary to establish a nondestructive testing system of fruit sweetness,improve the classification of fruit sweetness,and ensure the internal quality of fruit.After apple picking,the consumption of internal material and energy affects the internal quality of apple,which can be expressed by the aroma of apple itself.Aiming at the nondestructive testing of apple sweetness,this paper proposes a new odor recognition system based on Zig Bee wireless transmission network,which aims to establish the relationship between apple odor and sweetness and realize the nondestructive testing of apple sweetness.It is found that the gas concentration emitted by apples with different sweetness is different,which can be used as an important factor to distinguish apples with different sweetness.The greater the difference in sweetness,the more obvious the gas concentration.Therefore,apple sweetness is divided into five ranges,namely [12.0%-13.0%],[13.0%-14.0%],[14.0%-15.0%],[15.0%-16.0%],[16.0%-17.0%].Firstly,the experimental environment needed for data acquisition is set up,and the sensor array is used to collect the smell of apples in a closed bottle,and the collected gas data is sent to the PC through the wireless transmission system.Secondly,the data with abnormal values and noise influences are preprocessed,and then the eigenvalues are extracted to construct a good neural network model.Finally,the BP neural network model and CPSO-BP network model are used to classify and identify apples with different sweetness intervals.The experimental conclusions certificate that the classification recognition rate of BP neural network model is 76.67%,that of CPSO-BP neural network model is 83.33%,and the detection accuracy is improved by 6.66%,which is equivalent to the detection accuracy of commercial near infrared spectrum analyzer on instruments and equipment,and can realize nondestructive detection of apple sweetness.