Non-destructive Detection Of Tea Stalk And Insect Foreign Body Banse On Terahertz Time-domain Spectroscopy And Image

At present,the food industry is plagued by the problem of effective detection of lowdensity foreign bodies,and the current research on the detection of low-density foreign bodies in the food field for terahertz is still in its infancy,and many studies are still limited to laboratory conditions,such as high accuracy but time-consuming and laborious detection methods,or fast and simple detection methods but low detection success rate,and have not taken a solid step on the road to real-time detection application.Therefore,in this paper,two representative foreign bodies,endogenous foreign body and exogenous foreign body,terahertz food low-density foreign body,were selected to study from the aspects of spectrum and image,and explore the possibility of combining the atlas for low-density foreign body detection.(1)By modeling the spectral time domain signal and absorption coefficient as input,the optimal result obtained is generated by directly applying terahertz time domain signal as input to establish a KNN model,without using pretreatment method,the accuracy of identifying insects and tea stem substances mixed into tea leaves is 100%,the accuracy is 95.6%,and the recall rate is 98.7%.In the absence of a significant peak in the absorption coefficient,the terahertz time domain signal can achieve its foreign body recognition effect well,and the operation is simple,without the need to measure and enter sample thickness parameters.(2)Compared with terahertz time domain spectroscopy,terahertz time domain images provide a large-range detection method.Electromagnetic vibration feeding separates the sample from foreign matter,improving sample stacking.Traditional terahertz image acquisition is a full-band image with a long acquisition time and cannot achieve real-time rapid detection of low-density foreign objects.In contrast,the advantages of short acquisition time and low acquisition cost of terahertz single-frequency images reflect that the terahertz single-frequency images have the potential to replace the terahertz full-band data acquisition images to identify tea foreign bodies,indicating that electromagnetic vibration feeding-assisted terahertz timedomain signal technology is feasible in detecting low-density foreign bodies in tea stems and insects.This study verifies the feasibility of real-time detection of foreign bodies in food,and provides a new perspective for the application of real-time detection of foreign bodies in the food field.