Abstract: Gear is a common rotating mechanical part in machinery and equipment, and its fault diagnosis is an important means to avoid major accidents, and gear fault diagnosis is bound to become the focus of scholars' research. Machine vision technology provides a new idea for gear fault diagnosis. This paper analyzes the basic principle of machine vision system is analyzed, and the research status of gear image collection, gear image processing and gear image recognition based on machine vision is reviewed, the main problems of gear fault diagnosis based on machine vision are summarized. Some suggestions are put forward, such as the engineering research and application of dynamic gear image acquisition, three-dimensional gear image acquisition, the application of image enhancement and transfer learning technology, and the fusion of shallow learning and deep learning. It provides reference for technical personnel engaged in gear fault diagnosis related fields.