Hydraulic structure has a long history, which varies in many kinds. Its damage will have a great impact on people’s life and property safety. Therefore, relevant personnel should be able to immediately discover the damage of hydraulic buildings and focus on it, and at the same time, it can immediately carry out relevant warning and treatment of the safety risks. The traditional displacement measurements are mainly contact, which have lots of disadvantages, such as insufficient coverage area, many monitoring blind areas and high monitoring cost. With the rapid development of high definition camera and artificial intelligence, machine vision is a non-contact, high precision, low cost, all-weather monitoring system has received more and more attention. The machine vision system can not only eliminate the shortcomings of the contact measurement method, but also realize the multi-azimuth full coverage monitoring of hydraulic buildings. This paper will from the machine vision system composition and commonly used computer vision methods, matters to pay attention to three aspects of the current machine vision-based hydraulic building displacement monitoring methods are discussed, aimed at relevant researchers through machine vision to improve the monitoring of comprehensiveness, accuracy and real-time, help to better protect the health, safety, smooth operation of hydraulic buildings.