(北京工商大学 人工智能学院/食品安全大数据技术北京市重点实验室, 北京 100048)
(School of Artificial Intelligence/Beijing Key Laboratory of Big Data Technology for Food Safety, Beijing Technology and Business University, Beijing 100048, China)
In order to solve the problems of high cost of model maintenance, insufficient model scalability and poor sharing of spectral resources in spectral detection technology, and to improve the limitations of detection and analysis objects of dedicated portable spectrometers, a real-time online spectral detection system consisting of portable spectrometers, mobile terminals and cloud servers was developed using Android development tools and Java language by decoupling the spectral data collection and analysis functions. Based on cloud computing services, the establishment and analysis of cloud spectral models were completed. Taking the quantitative analysis of wheat flour gluten as an example, the quantitative analysis model of wheat flour gluten was established by using multiple scattering correction, competitive adaptive reweighted sampling method, and partial least squares regression algorithm. 25 wheat flour samples were tested and the analysis results were returned to verify the reliability of the system. The results showed that the error range of the gluten content of 25 wheat flour samples was 0-0.7%, and the average analysis time was 7.09s. The error range of gluten content and the time consumption of analysis results were both within the acceptable range, which verified that it was feasible to realize real-time online detection and analysis of spectra based on cloud computing services. By deploying quantitative analysis models of the main components of different foods to the cloud, the system could realize real-time online detection and analysis of various food qualities, hoping to provide technical support for rapid and non-destructive testing of food quality in multiple scenarios.