Recognition of Plasma-Treated Rice Based on 3D Deep Residual Network with Attention Mechanism

Abstract

Low-temperature plasma is a new agricultural green technology, which can improve the yield and quality of rice. How to identify the harvest rice grown by plasma seed treatment plays an important role in the popularization and application of low-temperature plasma in agriculture. This study collected hyperspectral data of harvest rice, including plasma seed treated rice, and constructed a recognition model based on the hyperspectral image (HSI) by 3D ResNet (HSI-3DResNet), which extracts spatial spectral features of HSI data cubes through 3D convolution. In addition, a spectral channels 3D attention module (C3DAM) is proposed, which can extract key features of spectra. Experiments showed that the proposed C3DAM can improve the recognition accuracy of the model to 4.2%, while the size and parameters of the model only increase by 4.1% and 3.8%, respectively. The HSI-3DResNet proposed in this study is superior to other methods with the overall accuracy of 97.47%. At the same time, the algorithm proposed in this paper was also verified on a public dataset.

TRiP: a transfer learning based rice disease phenotype recognition platform using SENet and microservices

“…This attention mechanism’s dynamism and adaptability lead to a marked enhancement in performance compared to conventional models. Its strength lies in its capacity to zoom into specific input components based on their relevance to the current query ( Tang et al., 2023 ).…”