Sparse/Feature-based VSLAM

Parent: Visual SLAM Implementation Framework ,  slam_index See also: Feature-based vs direct SLAM workflow

Source: cometlabs

• Front-end part of the Visual SLAM Implementation Framework
• Use only a small selected subset of the pixels in an image frame
• Feature maps generated are point clouds –> used to track the camera pose
• Requires feature extraction and matching
• To minimise: reprojection error (difference between a point’s tracked location and where it is expected to be given camera pose estimate)
• Pose estimation based on RANSAC
• A frame with most of its features concentrated in a small area: bad as the features are more likely to overlap

Sparse

Semi-dense