3D Motion Prediction in Multi-Domain Application Scenarios: Methods, Challenges and Future Directions

Abstract

3D motion prediction and tracking are the core technology of intelligent perception in the field of computer vision. It is widely used in the fields of complex environment perception of autonomous driving, industrial production safety monitoring, digital protection of intangible cultural heritage dance movements and so on. In recent years, the achievements in this field have emerged, but the existing research focuses on the customization and optimization of a single scene, and there are problems such as the separation of technical context, the disconnection between theoretical innovation and engineering implementation. The relevant review also has the defects of single coverage, lack of systematicness, and incomplete coverage of the latest frontier achievements. This paper takes 3D motion prediction and tracking technology in multi-domain application scenarios as the research object, systematically combs the core research results at home and abroad from 2017 to 2026, constructs a four-dimensional analysis framework, clarifies the technological evolution context, compares the applicable boundaries of different technical routes, and refines the adaptation rules of technology and scenarios, which can provide reference for relevant theoretical research and engineering application.