Nowadays, more and more projects and research require people to deal with getting information about people's faces. However, there is no chance to continue the work without knowing classifying organisms on the face, like which parts are eyes and which part is nose. Therefore, face alignment, a method of localizing these facial points, appears to be important. The problem of face alignment can be solved in several ways. Though traditional methods under constrained environments can solve problem of face alignment effectively, there are numerous techniques developed over the last two decades for face alignment, including traditional methods and deep learning-based methods. This paper describes the development process and basic methods of face alignment technology, which are methods based on traditional learning represented by ASM and AAM and deep learning methods represented by CPR and Heatmap. In the end, this paper points out that there are more challenges to facing alignment without constraints in the future. Peoples need to provide more relevant datasets, evaluate metrics and identify the accuracy of different algorithms applied in data sets.