基于改进ResNet模型的食品新鲜度识别方法

目的:解决现有食品新鲜度识别方法存在的检测效率低和精度差等问题。方法:基于食品生产线图像采集系统，提出一种改进的残差神经网络模型用于生产线食品新鲜度识别。引入改进的LRELU激活函数提高模型的识别性能，引入批量归一化层提高模型的训练效率，引入Dropout层丢弃一定比例的神经元降低过拟合的影响。结果:与常规食品新鲜度识别方法相比，试验方法能够较为准确、高效地实现食品新鲜度识别，总体新鲜度识别准确率>97%,平均识别时间为9.8 ms,满足食品生产线对新鲜度识别的需要。结论:基于深度学习的检测方法是一种无损、高效、高精度的食品图像新鲜度识别方法。

Food freshness recognition method based on improved ResNet model

Objective: Solve the problems of low detection efficiency and poor accuracy in existing food freshness recognition methods. Methods: Based on the food production line image acquisition system, an improved residual neural network model was proposed for food freshness recognition on the production line. The improved LRELU activation function was introduced to improve the recognition performance of the model, the batch normalization layer was introduced to improve the training efficiency of the model, and the Dropout layer was introduced to discard a certain proportion of neurons to reduce the impact of over fitting. Results: Compared with conventional food freshness recognition methods, the experimental method could accurately and efficiently achieve food freshness recognition, with an overall freshness recognition accuracy of >97%, average recognition time of 9.8 ms, which meet the needs of food production lines for freshness recognition. Conclusion: The detection method based on deep learning is a non-destructive, efficient, and high-precision method for recognizing the freshness of food images.