A Review of the development of Visual SLAM

Simulated Localization and Mapping (SLAM) technology is one of the core technologies in the navigation field of autonomous mobile robots and unmanned aerial vehicles, which enables the machines to simultaneously locate themselves and build environmental maps through visual sensors (such as cameras). This technology has important applications in the fields of self-driving cars, augmented reality, and automatic navigation. This paper aims to provide a comprehensive overview of visual SLAM, focusing on its theoretical basis, key technologies, and development in the past decades. This paper first summarizes the basic principle and classification of SLAM technology, and then deeply discusses the theoretical basis of visual SLAM, including image processing, computer vision, state estimation, and kinematics. Then, the paper introduces the development of key technologies of visual SLAM in detail. Finally, this paper discusses the application of visual SLAM in UAV navigation, robot exploration, and virtual reality, and points out the limitations of current technology and the future research direction, including multi-sensor fusion, deep learning application, real-time performance improvement and SLAM technology in antagonistic environment. The main contribution of this review is to sort out the latest progress of visual SLAM in detail and look forward to the future development trend and potential research direction.