Nondestructive Testing Of Processing Tomato Fruit Quality Based On Hyperspectral Imaging Technology

【Objective】Tomato is the second largest vegetable crop in the world,and processing tomato is a cultivated type of tomato,with more than 20 million tons of processing tomatoes harvested annually for industrial needs worldwide.The selection of high quality processing tomato fruits as raw materials is the key to improving the quality of tomato products.About 10-20%of immature fruits mixed with processing tomatoes during mechanized harvesting need to be selected or manually screened,but traditional methods for quality and ripeness detection are time-consuming and laborious.The aim of this study is to develop a rapid and nondestructive detection model to identify the quality and ripeness of processing tomatoes,which can be used to screen high-quality processing tomato raw materials,thereby improving the quality of tomato products and increasing international competitiveness【Methods】In this paper,we used near-infrared hyperspectral imaging to detect the quality of processing tomatoes,processed the raw spectral data using Savitzky-Golay convolutional smoothing method,and developed regression and discriminant models of recurrent neural network(RNN),support vector machine(SVM),K-nearest neighbor(KNN),random forest(RF)and partial least squares(PLS)for the detection of hardness,titratable acid content,soluble solids content(SSC),lycopene content,overall quality and ripeness,to compare the effectiveness of different modeling approaches for quality prediction of processing tomatoes and to select the most suitable model for quality detection of processing tomatoes.【Results】1.In the study of nondestructive testing of the quality of processing tomatoes by near infrared hyperspectral imaging technology,the deep learning model RNN is superior to the other four machine learning models in predicting the quality of processing tomatoes.R_P~2and RMSEP for the hardness detection model were 0.918 and 0.84,and RPD was 3.49.R_P~2and RMSEP for the SSC detection model were 0.877 and 0.19,and RPD was 2.85.R_P~2and RMSEP for the titratable acid assay model were 0.869and 0.03,and RPD was 2.76.The R_P~2and RMSEP for the lycopene assay model were 0.937 and 0.73,and the RPD was 3.98.2.In the study of ripeness discrimination detection of processing tomatoes.The R_P~2and RMSEP of the RNN model were 0.649 and 0.56,and the RPD was 1.69.The prediction accuracy of this model was lower than that of the single quality index prediction results.3.In the study of ripeness discrimination detection in processing tomatoes,compared with other machine learning models,the deep learning model RNN performed best in the discrimination of ripeness of processing tomatoes,and the accuracy of this classifier for ripeness discrimination of processing tomatoes was The classification accuracy of the RNN model was 17%higher than that of the PLS-DA model and64%higher than that of the KNN model.4.Based on the above research,a non-destructive testing system for the internal quality and ripeness of processing tomatoes is designed,which is used to test the hardness,titratable acid,lycopene,SSC,and ripeness of processing tomatoes.The whole system consists of two parts:the front-end page can be used to design the interface for quality testing and divide different functional areas,and the back-end module is the database for the quality testing of processing tomatoes,which contains encapsulated sub-modules such as pre-processing and models,and the user can call the database of the back-end module through the designed front-end page to achieve rapid testing of multiple qualities of processing tomatoes.【Conclusion】In summary,the deep learning model RNN performs well in both internal quality detection and ripeness discrimination of processing tomatoes,proving the possibility of RNN models to predict the internal quality and ripeness of processing tomatoes,providing new ideas for nondestructive quality detection of other agricultural products,establishing a quality detection system for processing tomatoes,and providing technical support for harvesting and grading of processing tomatoes.