Study On Quality Prediction Of Pumpkin During Microwave Drying Based On Multi-Source Information Fusion

Pumpkin contains various physiological active substances.In China,the annual production of pumpkins is enormous,accounting for about 30%of the total production in the world.Therefore,pumpkin has important medicinal and economic value.The number of pumpkins directly consumed after picking is very small.Most pumpkins are processed into dried products through drying.The purpose is to extend the shelf life of pumpkins.Microwave drying technology has the advantages of high efficiency,energy saving and quality improvement,which is widely used in the development of pumpkin dry products.With the development of technology,microwave drying technology is gradually becoming intelligent by combining non-destructive testing technologies such as machine vision and electronic nose,which can continuously meet the comprehensive evaluation needs of sensory and flavor of dried products,thus avoiding the subjectivity of traditional manual evaluation.However,when multiple detection sources or sensors such as machine vision and electronic nose are included in the system at the same time,the obtained information may have complex relationships such as redundancy and contradiction,leading to a decrease in the accuracy of quality evaluation.In response to the above issues,this study focuses on pumpkin as the research object,using machine vision and electronic nose as the main detection tools to explore the changes in drying characteristics,structure,and quality of pumpkin during microwave drying.A single source and multi-source information fusion prediction model based on visual information and odor information was established,which can provide reference for the optimization of pumpkin dried product processing technology and quality prediction research.The main research content includes:(1)In order to satisfy the experimental conditions of pumpkin microwave drying and the requirements of multi-source information collection,a multi-source information collection microwave drying experimental system has been designed and built,which can achieve constant temperature microwave drying of pumpkin and online collection of quality information,visual information,and odor information during the drying process.(2)This study investigated the changes in drying characteristics,structure,and quality of pumpkin during microwave drying.The results showed that the microwave drying process of pumpkin can be divided into three stages:Rising Rate Drying Period(RRP),Fast Falling Rate Drying Period(FFRP),and Slow Falling Rate Drying Period(SFRP);As the water content decreases,the bulk density gradually decreases,the micro porosity gradually increases,the rehydration ratio gradually increases,and the content ofβ-carotene and vitamin C gradually decreases;At different temperatures(60℃,70℃,80℃),the water content,β-carotene,and vitamin C follow the zero order,third order,and first order degradation kinetics,respectively,with prediction accuracy R~2greater than 0.96,0.93,and 0.95.(3)This study investigated the feasibility of using a single source information prediction model to predict the drying characteristics,structure,and quality of pumpkin during microwave drying.The research results showed that the prediction accuracy of the visual information based quality prediction model for moisture content,bulk density,andβ-carotene during the drying process was in the range of 0.9704～0.9843,0.9841～0.9910,and0.9781～0.9858,respectively,with good prediction results;The prediction accuracy of the quality prediction model based on odor information for moisture content,bulk density,andβ-carotene during the drying process is in the range of 0.7535～0.9478,0.8473～0.9083,and0.9050～0.9460,respectively,with good prediction results.(4)This study investigated the feasibility of multi-source information fusion prediction models for predicting the drying characteristics,structure,and quality of pumpkin during microwave drying.The research results showed that multi-source information fusion prediction models were established using Extreme Learning Machine(ELM),Partial Least Squares Regression(PLSR),and Back propagation neural network(BPNN),respectively,The prediction accuracy of moisture content,bulk density,andβ-carotene during pumpkin microwave drying process is in the range of 0.9758～0.9857,0.9861～0.9926,and0.9850～0.9874,respectively.Compared with the single source information model,the prediction accuracy is improved,indicating that multi-source information fusion technology can be used to improve the prediction accuracy of drying characteristics,structure,and quality during pumpkin microwave drying process.