With the further deepening of world trade, the demand for English learning is growing. The Internet teaching system provides a broader way for English learning. Teaching resources include pictures, videos and other forms. Based on the actual needs of business English, this paper designs an Internet business English teaching system combining machine learning algorithms and high-resolution graphic recognition methods. Based on the B/S three-tier structure, the system includes three main structures: system management, video course management, and system learning function realization, which meet the resource sharing and expansion requirements of the English teaching video system. As the multimedia teaching resources on the teaching platform mainly consist of pictures and videos and students' learning data are rich, this paper uses the hybrid collaborative filtering algorithm to analyze and integrate the data information collected by the teaching system, so as to realize the analysis of students' learning, and further realize personalized teaching. At the same time, the algorithm of convolutional neural network is applied to the image high resolution process of the video teaching system to solve the recognition error caused by the pixel problem of the image. By using interpolation algorithm to reconstruct the image with high resolution, the efficiency of the algorithm is improved and the resources of the teaching system are integrated. Finally, the overall performance of the video teaching system is tested to prove the effectiveness and usability of the model in this paper. Based on this, this paper puts forward some suggestions on the interactive strategies of online business English teaching in the context of the Internet.