Research On Intelligent Detection Of Apple Freshness Based On Low-field Nuclear Magnetic Resonance Technology

Apples have high nutritional value and are popular among the public.As one of the mainstream fruits in Chinese fruit market,the issue of freshness and quality has always been a concern.During the storage process,the growth of microorganisms,evaporation and loss of water can lead to a decline in quality.Research on apple freshness detection methods can help consumers better understand and choose products.Traditional freshness identification mainly relies on sensory evaluation and physicochemical detection.The judgment criteria for sensory evaluation are difficult to unify and subjective.Although the results of physicochemical detection are more objective,it requires professional testers and has the characteristics of long detection cycles and complicated procedures.Moreover,it is difficult to establish a relationship between various indicators,making it impossible to provide a fast,comprehensive,and accurate evaluation of apple quality.With the development of modern agricultural technology,intelligent detection has become a research hotspot in the fruit and vegetable market.Research on intelligent detection of fruit freshness has important significance in promoting the development of preservation technology and improving economic benefits.Therefore,based on low-field nuclear magnetic resonance detection technology,this study investigated the correlation between the moisture status and physicochemical quality indicators of five apple varieties(Tianshui Huaniu,Aksu Bingtangxin,Huangyuan Shuai,Luochuan Red Fuji,and Cream Fuji)at different storage periods.Quantitative methods such as partial least squares regression and support vector machines were used to construct detection models for apple freshness and quality indicators.Based on the C#language platform,an intelligent detection system for apple freshness was developed,which achieved the detection of apple freshness and related physicochemical quality indicators.The main research content and specific conclusions of this article are as follows:1.Within a storage period of 25 days,the moisture content,hardness,color difference,soluble solid content,titratable acidity,and moisture status of apples were measured every 5 days to explore the correlation between apple moisture status and freshness and quality indicators.The results showed that:1)apple freshness was classified into fresh,slightly fresh,acceptable,and rotten based on sensory evaluation and cluster analysis.2)The moisture content,hardness,color difference,soluble solid content,and titratable acidity of the five apple varieties changed significantly during the entire storage period.The Fisher discriminant analysis model established based on physicochemical indicators as discriminant factors can effectively identify the freshness of apples.3)According to the correlation analysis results,there is a significant or extremely significant correlation between sensory evaluation scores and related quality indicators and moisture status throughout the storage period.Therefore,the results can provide a basis for lowfield nuclear magnetic resonance-based detection of apple freshness and related quality.2.Based on the correlation between apple quality indicators and moisture status,support vector machines and partial least squares regression were used to construct apple freshness classification and related quality prediction models based on low-field nuclear magnetic resonance detection technology.The detection performance of the two models was compared,and the results showed that:1)both models can accurately and effectively identify the freshness of Tianshui Huaniu apples,and the support vector machine model is better.2)In the quantitative prediction models of quality indicators during storage,the partial least squares regression model has excellent prediction ability for titratable acidity and hardness during storage,and good prediction ability for moisture content,soluble solid content,and color difference,but its ability to predict color difference is not as good as that for moisture content and soluble solid content.The support vector machine model has excellent prediction ability for titratable acidity,moisture content,and soluble solid content during storage,good prediction ability for hardness,and moderate prediction ability for color difference.3)Without considering the differences in apple varieties,using all apples as test samples,the recognition rate of the partial least squares regression model for freshness is 87.50%,and that of the support vector machine model is 92.50%.Both models can accurately and effectively identify the freshness of apples,and the support vector machine model is better.4)Without considering the differences in apple varieties,using all apples as test samples,in the quantitative prediction models of quality indicators during storage,the partial least squares regression model has excellent prediction ability for hardness and moisture content during storage,good prediction ability for color difference and soluble solid content,and moderate prediction ability for titratable acidity.The support vector machine model has excellent prediction ability for moisture content and soluble solid content during storage,good prediction ability for titratable acidity and hardness,and moderate prediction ability for color difference.The above results show that both models have excellent prediction effects on moisture content during storage.Therefore,using low-field nuclear magnetic resonance detection combined with partial least squares regression and support vector machine models can effectively predict the freshness and related quality of apples.3.Based on the established detection models,an "Apple Freshness Intelligent Detection System" was developed using the C#language.This system can manually input real-time or batchimport low-field nuclear magnetic resonance signal data,and intelligently detect the freshness and related quality indicators of multiple varieties of apples.In addition,this system has an upgrade and optimization function,which can expand the model database to achieve intelligent detection of more fruits.The system has a simple and friendly interface,easy operation,clear and intuitive detection results,and has realized the rapid intelligent detection of apple freshness.