SLAM (Simultaneously Localizing and Mapping) is a machine vision concept that is widely used in robotics, and Visual SLAM is a type that uses the camera as a primary sensor to create maps from captured images. Mohammad Javadian Farzaneh wrote an article on Towards AI to explain the purpose and workings of Visual SLAM to aid those who are new to the concept. One of the keys to Visual SLAM is extracting feature points from images to create a map, and reliable algorithms such as SIFT, SURF, and ORB are used for this purpose. The process also involves image matching, estimation, and triangulation to find the camera’s trajectory and the real-world position of feature points.

source update: Visual SLAM, A Booster Overview – Towards AI