Abstract: Grain and oil safety is one of the important food safety issues and has received widespread attention worldwide. Therefore, rapid, accurate and efficient detection technologies are crucial for ensuring the safety of grains and oils. However, traditional detection methods for grains and oils have disadvantages such as long-time consumption, large subjective errors and poor real-time performance, which cannot meet consumers’ high requirements for food quality. The combination of computer vision and deep learning provides a rapid, efficient and non-destructive solution for grain and oil detection. This article introduces the basic principles of deep learning and computer vision and their advantages in food detection, focusing on the application of algorithms such as convolutional neural network (CNN), long short-term memory (LSTM), and generative adversarial network (GAN) in grain and oil detection. Meanwhile, it demonstrates the significant effects of these technologies in improving the detection accuracy and efficiency and summarizes recent progress in the application of computer vision and deep learning in non-destructive testing of grains, oils and their products. Finally, the limitations and future trends of computer vision and deep learning in the field of grain and oil safety are discussed from various aspects such as optimizing the robustness and interpretability of the model and developing lightweight models to adapt to the resource-constrained detection environment, aiming to promote the development of more efficient and accurate food detection technologies.

Key words: computer vision; deep learning; grains and oils; non-destructive testing