Aiming at problem that offset even lost will occur when illumination changes drastically or meets with rotation in plane and a large block

a spatio-temporal context learning tracking algorithm is proposed based on locality sensitive histogram. The algorithm is based on Bayesian framework and use biological visual characteristics based on locality sensitive histogram

extract illumination invariant features

combined with low-level gray scale features

set up statistics related model for target and background to realize tracking. Extensive experimental results show that the proposed algorithm performs good tracking effect and central error is small and robustness is strong compared with STC and MIL algorithms.