Abstract: Crop pests threat our country's food security, crop pest control is a key link to ensure food security, and crop pest detection based on machine learning is an important method for pest detection．Characteristics of publicly available crop pest datasets has been introduced firstly, and application of traditional machine learning methods such as K-means, SVM, and BP networks in crop pest detection was discussed deeply．Traditional machine learning methods were easily interfered by image noise in feature extraction and detection accuracy, especially to detect crop pest categories in complex images．Usage of deep learning methods such as Faster RCNN, SSD, and YOLO in crop pest detection was further introduced．Methods based on deep learning had stronger feature extraction capabilities and could detect crop pest targets in complex images with higher accuracy and speed．Currently, crop pest detection model based on deep learning still needs further exploration in areas such as detecting small targets, real-time processing, densely distributed pests, and pests against complex backgrounds, to provide technical support for intelligent control of agricultural pests．