Visual slam method and apparatus based on point and line features

Abstract

Systems and methods of visual simultaneous localization and mapping (SLAM) based on point and line features are disclosed. The method includes: receiving a visual image frame input by a camera device; extracting feature points and feature lines from the visual image frame; predicting a first pose of the camera device using the feature points; observing an extracted first feature line to determine a feature line measurement of the first feature line; obtaining a global feature line state vector set of the camera device, where the global feature line state vector set includes feature line state vectors of N historical feature lines, and N is a positive integer; and updating the first pose using the feature line measurement and the global feature line state vector set. In this way, feature point-based motion estimations and observation features of feature lines observed in an environment are integrated to update a pose of the camera device in real time, thereby improving visual SLAM accuracy.