Detecting Tomato Internal Quality Based On Hyperspectral Imaging Technology

With the rapid development of the global economy and the general improvement of people's living standards,consumers pay more attention to the internal quality,storage and transportation quality and excellent flavor quality of tomatoes.The traditional method of fruit quality detection is time-consuming and destructive,which is difficult to apply to the batch detection of fruits.As a non-destructive testing method,spectral technology is widely used in fruit detection.Hyperspectral imaging technology can detect the two-dimensional space set and one-dimensional spectrum information of the target at the same time.By combining the image and spectral characteristics,the overall spatial spectral information of the tomato can be obtained,which has obvious advantages in the non-destructive detection of tomato internal quality.Visible/near-infrared spectroscopy technology has the advantages of being fast,convenient and low in detection cost.The internal quality of tomatoes and cherry tomatoes was detected based on hyperspectral imaging technology and visible/near infrared spectroscopy technology in this study.Successive projections algorithm?SPA?,interval random frog?i RF?,improved interval random frog?mi RF?,uninformative variable elimination?UVE?,principal component analysis?PCA?and successive projections algorithm combined with principal component analysis?SPA-PCA?were used to extracts the characteristic wavelength.Partial least squares regression?PLSR?and extreme learning machine?ELM?models were established based on the characteristic wavelength to detect the soluble solids content?SSC?,hardness,comprehensive quality of tomatoes.In addition,the effects of different characteristic wavelength selection methods on model detection accuracy were compared.The main research contents and conclusions of this article are as follows:?1?Research on non-destructive determination of tomato SSC and hardness based on hyperspectral imaging technology,the influence of characteristic wavelength selection methods on the detection accuracy was studied.Established a detection model based on PLSR,and comparatively analyzed the detection results of SPA-PLSR,mi RF-PLSR,i RF-PLSR,UVE-PLSR models.The mi RF algorithm effectively constructs the initial variable subset based on the i RF.The results show that mi RF effectively reduces the complexity of the model and improves the timeliness and accuracy of the tomato quality detection model compared with i RF.Among the four models,the PLSR model based on the mi RF has the best detection effect on tomato SSC and hardness.The test set correlation coefficient?R_p?is 0.8725 and0.9036,respectively.And the test set root mean square error?RMSEP?is 0.2973°Brix and0.006 kg·mm^-2,respectively.?2?Research on non-destructive determination of cherry tomato SSC based on hyperspectral imaging technology,the performance of the PLSR model based on different characteristic wavelength selection methods was studied.The results of PCA-PLSR,SPA-PCA-PLSR and mi RF-PLSR models were analyzed and compared.The SPA-PCA algorithm uses SPA and PCA in series to improve the effectiveness of the characteristic band.The research shows that the prediction accuracy of the model built based on the principal component extracted by the SPA-PCA has been significantly optimized.Among the three models,the SPA-PCA-PLSR model has the best detection effect,with R_p 0.9039 and RMSEP0.5582°Brix.?3?Research on non-destructive determination of cherry tomato comprehensive quality based on sensory evaluation.Visible/Near-infrared spectroscopy technology was adopted in this study.The comprehensive quality discrimination index of cherry tomato was established by combining physical and chemical components and sensory evaluation,and the detection model of cherry tomato comprehensive quality was established based on PLSR and ELM,and the influence of different modeling methods on the detection results was compared.The results showed that SPA-PCA-PLSR model had the best detection results,with R_p 0.7715 and RMSEP 0.0683,which indicated that it was feasible to detect the comprehensive quality of cherry tomato by fusion sensory evaluation.In addition,the performance of PLSR model is better than that of ELM model in this study.