Abstract: Grain storage security is an important issue to ensure national food security. In the process of grain storage, the monitoring of grain pests is very important. Common methods for detecting grain pests mainly include manual detection, sound detection, traditional machine learning image processing, and deep learning image processing. The traditional machine learning image processing includes image preprocessing, feature extraction, and feature classification. Image preprocessing uses denoising, filtering and image compression technology to improve the image recognition and compress the image size. Feature extraction refers to the extraction of morphological features and color features of grain insects. In feature classification, various algorithms are used to optimize SVM classifier and other classifiers to complete the classification of grain pests. The deep learning technology includes image classification and object detection. The two-stages and one-stage object detection algorithms for grain pest detection applications are introduced in detail. The two-stages object detection is mainly improved based on Faster R-CNN and R-FCN algorithms, and the one-stage target detection uses SSD algorithm and other designed end-to-end algorithms. In order to improve the robustness of the algorithm, the generation countermeasure network is used to improve the classification effect. Then, the methods of estimating the population density of stored grain pests after using image detection technology are introduced, including using video surveillance, special classifiers and other means to judge the live insects and dead insects, and even judge the population density of stored grain pests. In this review, the state-of-the-art research methods of grain pest detection are reviewed, and the future development of image processing technology in grain pest detection is prospected.