Passive Initialization Method Based on Motion Characteristics for Monocular SLAM

Algorithm 2

Error reduction with subsequent BA.

Require: Frame I: Initialization Frame

Frame n: Subsequent Frame

: Feature points of Frame I

: Feature points of previous frames corresponding to

: Optimized rotation matrices of previous frames

: Optimized translation vectors of previous frames

: Initialized Map Points

Ensure: Map points and Initialization Evaluator .

Extract feature points of Frame n and match with to obtain matched feature points .

Optimize the Reprojection Error with Subsequent BA and thereby obtain optimized , , .

Calculate current evaluator value .

ifthen

Stop iteration, and return , , as initialization results.

ifis too big then

Stop iteration, and report failure of initialization.