The purpose of this article is to explore the preview behavior of junior high school mathematics average students in the distance education environment, and make an in-depth analysis by using image processing technology. Aiming at the problem that students’ learning behavior in distance education is difficult to observe directly, a preview behavior analysis method based on image processing is proposed. In this study, a data set containing the preview behavior images of junior high school mathematics average students is constructed, and then the convolutional neural network (CNN) is used for feature extraction and behavior classification to identify students’ different preview behaviors. The results show that CNN model can accurately identify students’ preview behavior categories and achieve high classification accuracy. Recurrent neural network (RNN) model also successfully predicted the students’ future preview behavior trend, showing good prediction performance. By comparing different feature extraction methods, the advantages of CNN feature extraction are verified. The preview behavior analysis method based on image processing is feasible, which provides a new perspective for students’ behavior analysis in the distance education environment.