Fast traceability detection of Astragalus membranaceus based on the combination of electronic tongue and electronic eye to improve MobileNetv3

Abstract

Objective: To realize the rapid traceability detection of Astragalus membranaceus from different origins. Methods: This study proposed a rapid detection method for the origin of Astragalus membranaceus based on the improved MobileNetv3 network based on the combination of electronic tongue and electronic eye. The electronic tongue and electronic eye were used to collect the one-dimensional fingerprint and two-dimensional appearance image information of different samples of Astragalus membranaceus. The Gramian Angular Field (GAF) was used to convert the one-dimensional electronic tongue signal into two-dimensional image information, retain the time series related features in the electronic tongue signal, and then fused them with the image information collected by the electronic eye. Finally, the MobileNetv3 model improved based on Pyramid Split Attention (PSA) was adopted to realize the classification and recognition of Astragalus samples from different habitats. Results: The experimental results showed that the method in this paper had higher recognition accuracy than using electronic tongue or electronic eye alone. The accuracy, precision, rrecall and F1-score of the test set were 98.8%, 98.8%, 98.8% and 0.99, respectively. The classification accuracy of the improved MobileNetv3 network was 8% higher than that of the original model, and the parameter quantity was only about 1/5 of the original parameter quantity. Conclusion: The improved MobileNetv3 network can effectively reduce the calculation of parameters and improve the recognition accuracy of Astragalus membranaceus from different origins.

Keywords

• electronic tongue
• electronic eye
• gramian angular field
• data fusion
• mobilenetv3
• astragalus membranaceus