An efficient multiscale integrated attention method combined with hyperspectral system to identify the quality of rice with different storage periods and humidity

TL;DR: This study proposes a multiscale integrated attention method (MSIA) combined with hyperspectral technology to identify rice quality with varying storage periods and humidity, achieving 99.69% accuracy and outperforming other spectral recognition algorithms.

Abstract: As an essential food, the quality of rice is critical to human health. The quality of rice varies significantly under different storage conditions. Using spectral analysis technology, a non-destructive testing method can quickly identify rice with quality, but an effective data analysis method directly affects its detection performance. Aiming at the characteristics of strong correlation and high dimensionality of rice spectral information, this paper proposes an effective feature extraction method to mine the spectral information for quality identification of rice with different storage periods and humidity. Firstly, the hyperspectral system is employed to obtain the spectral information of rice during six storage periods with different storage humidity. Secondly, a multiscale integrated attention block (MSIA) is proposed, which consists of multiscale feature extraction and spatial-channel attention mechanisms that can realize deep decoding of the spectral information. Finally, MSIA is combined with classical convolutional neural networks (CNN) to detect the quality of rice at different storage periods and humidity. The method's effectiveness is demonstrated in comparison with other advanced spectral information recognition algorithms. Specifically, the integration of ShuffleNetV1 and MSIA achieves the best overall classification performance with an accuracy of 99.69%, precision of 99.72%, recall of 99.69%, and F1-score of 99.69%. These results indicate that MSIA is an efficient spectral feature extraction technique. The combination of ShuffleNetV1 + MSIA and the hyperspectral system presents a novel approach for rice quality assessment with different storage periods and humidity.