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

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