A DualModeAttention (DMA) method is proposed, which adaptive captures the featured salient regions of rice disease
images in diversity of rice diseases and complex backgrounds by calculating the similarity weight corresponding to
each spatial position to improve the detection accuracy of rice models.
An Residual Adaptive Block(RAB) method is proposed, which solves the problem of fuzzy edges of rice images by
combining residual structure and channel attention to synthesize multiple spatial information, and performing multiple
dimensional changes with other fine adjustments to the feature maps.
A Particle Swarm Optimization Chameleon (PSOC) method is proposed. This method enhances the search capability
and convergence by combining particle swarm optimization with chameleon cluster algorithms, enabling particles to be
influenced by both their individual optimal positions and global optimal positions.