Author：Department of Electrical Engineering,NCHU/Electrical and Computer Engineering, NCTU/Department of Electrical Engineering,NTUT/Taiwan Agricultural Research Institute,COA

The quality control of agricultural products is a very important issue. Traditional methods use manual sampling. The disadvantage is that the procedures are complicated and time consuming, and the test results cannot be obtained immediately. The disadvantage of invasive inspection is that it destroys the test samples and makes them unable to be sold or eaten. Therefore, it is difficult to detect in large quantities or comprehensively, and cannot be widely used.
Hyperspectral detection technology, its spectrum is much richer information than general multi-spectral images, and can completely display the subtle differences in spectra of different targets. As long as the corresponding bands of various targets are found, the hyperspectral imaging technology can effectively classify them. And use artificial intelligence and machine learning models to automatically learn spectral features and characteristic bands, and apply the algorithm to the developed handheld hyperspectrometer to achieve the purpose of real-time detection and improve the traditional method of manual sampling.