In recent years, intelligent agriculture has shown great potential in improving the production efficiency and reducing the waste of resources. To solve the complex background and tobacco picking in the recognition accuracy and efficiency, the paper proposed a improved YOLOv5 of tobacco picking robot visual recognition model, the model combines binocular imaging technology and channel attention mechanism, at the same time introduce multi-scale features enhanced pyramid and instance normalization module to improve the generalization ability of the model. The experimental results show that the improved YOLOv5 model achieves 92% recognition accuracy, improves the accuracy to 90% in the complex background, and reduces the recognition time to 0.2 seconds, which is significantly better than the traditional model. The results show that the proposed model has higher utility and application potential in tobacco picking tasks and can provide efficient and precise solutions for automated tobacco picking.