A Hypothesis-Optimized RANSAC Algorithm for Track Initiation
Feng Ping Yang, Yujuan Luo, Qiang Li, Yuming Yin · 2020
Track initiation is still a challenge in extremely dense clutter environment. Since the validity of hypothesis in RANSAC approach can not be guaranteed in heavy clutter situations, a novel track initiation algorithm named hypothesis-optimized random sample consensus (HO-RANSAC) is proposed to address the above problem. The proposed HO-RANSAC uses hypothesis splitting and merging strategy in the hypothesis verification procedure to improve the probability of valid hypothesis, in addition, local optimization is applied to update the hypothesis model. Simulation results demonstrate that the proposed algorithm has superior performance compared with RANSAC, the modified logic-based algorithm, the modified Hough transform and DB-RANSAC.