【摘要】 将传统深度学习的CBAM注意力机制算法直接用于害虫分类，得到的精度不理想，主要原因是害虫个体小、害虫与背景颜色差异小。为此，提出一种适用害虫分类的改进CBAM注意力机制算法。该算法通过改进通道注意力模块，提高害虫的关键信息特征在特征图中的比重，以解决害虫与背景颜色差异小的问题；改进空间注意力模块，以解决害虫个体小、感知难的问题。在消融实验中得到的结果表明，该算法能够有效解决害虫个体小、害虫与背景颜色差异小等问题，使分类准确率得到提升，达到75.9%。更多还原

【Abstract】 The CBAM attention mechanism algorithm of traditional deep learning is directly used in the classification of agricultural pests, and the accuracy obtained is not ideal.The main reason is that the agricultural pests are small, the background is complex, and the color difference between pests and the background is small.To solve these problems, an improved CBAM attention mechanism algorithm suitable for the classification of agricultural pests is proposed.By designing a local channel attention module, the algorithm makes the key information features of pests account for a larger proportion in the feature map to solve the problem of small differences between agricultural pests and background colors.We design a local cross-space attention module to solve the problem of small individual pests and difficulty in perception.The results obtained in the ablation experiment show that the algorithm can effectively solve the problems of small insects and small background color differences, and the classification accuracy rate has also been improved to 75.9%.更多还原