Pest Image Recognition in Rice Using Convolutional Neural Networks

水稻作为全球重要的粮食作物之一，其产量和品质对食品安全和农业经济具有重大影响。为应对稻田虫害分类中遇到的样本不平衡和特征复杂性问题，提出了一种基于卷积神经网络(CNN)的识别方法。该方法整合了ResNet、VGG等经典的网络结构，并通过实施数据增强和迁移学习策略，有效提升了模型的泛化能力与分类精度。在数据预处理阶段，引入了旋转、缩放和平移等多样化的增强技术，增强了模型对复杂农田环境的适应能力。为了解决类别不平衡的问题，采用了类别权重调整，特别提升了模型在小样本类别上的性能。通过集成学习策略进一步优化了模型的表现，显著提高了分类精度和系统稳定性。实验结果显示，优化后的CNN模型在测试集上表现卓越，整体分类准确率高达98.23%，在具体类别如“rice leaf roller”和“asiatic rice borer”上的准确率分别为96.5%和95.6%。对于样本量较少的“grain spreader thrips”类别，模型同样展现了优异的识别能力。模型在测试集上的平均精确率、召回率及F1分数分别为96.48%，98.41%和97.26%，进一步验证了所提出模型的高效性和鲁棒性。Rice is a crucial food crop affecting food security and agriculture. This study presents a CNN-based method to address challenges in rice pest classification, such as sample imbalance and feature complexity. The method uses ResNet and VGG architecture, data augmentation, and transfer learning to improve model generalization and accuracy. Preprocessing includes rotation, scaling, and translation to adapt to varying field conditions. Class weight adjustment is used to handle imbalance, enhancing performance on minority classes. Ensemble learning further optimizes model performance. The optimized CNN achieved 98.23% accuracy on the test set, with 96.5% and 95.6% accuracy for “rice leaf roller” and “asiatic rice borer”, respectively. It also showed strong recognition for the “grain spreader thrips” class with fewer samples. The model demonstrated 96.48% precision, 98.41% recall, and 97.26% F1 score, confirming its efficiency and robustness.