关键词: 番茄果实检测, YOLOv5, 数据增强, 注意力机制, 轻量化网络

Abstract: In order to realize the fast identification and localization of tomato fruit, an improved YOLOv5 algorithm is proposed, which improves the generalization of the model by improving the data enhancement method, improves the inference speed by replacing the lightweight backbone network, increases the attention mechanism, enhances the feature extraction ability of the backbone network, and accelerates the computation by modifying the special fusion layer to improve the recognition ability of the target and reduce the computational cost. The results show that the average accuracy of the proposed method is 87.5% and the accuracy of mature tomatoes is 90.1%, which is 4.2% and 1.9% higher than that of the unmodified YOLOv5s model, respectively. And on the test set, the experimental inference detection speed has reached 101 frames per second.

Key words: tomato fruit detection, YOLOv5, data enhancement, attention mechanism, lightweight network