Automatic recognition method for soft packaged food based on improved YOLOv3 model

Objective: To solve the problems of poor recognition accuracy and low efficiency of existing automatic recognition methods for packaged food. Methods: Based on the automatic identification system of packaged food, an improved YOLOv3 model was proposed for the automatic identification of soft packaged food. The Kmeans++algorithm was introduced into the model to solve the problem of small target insensitivity, the Mish activation function was introduced into the model to improve the accuracy of recognition, and the attention mechanism Senet was introduced into the model to improve the ability of feature extraction. The performance of the recognition model was analyzed through experiments, and the superiority of the model was verified. Results: Compared with the conventional recognition methods, the proposed method can more accurately and efficiently realize the automatic recognition of flexible packaging food, the recognition accuracy rate was 95.40%, and the recognition efficiency was 23.80 f/s, which meets the needs of packaging food recognition. Conclusion: By optimizing the existing food recognition model, the performance of the recognition model can be effectively improved.