Complex background, especially when the object is similar to the background in color or the target gets blocked, can easily lead to tracking failure. Therefore, a fusion algorithm based on features confidence and similarity was proposed, it can adaptively adjust fusion strategy when occlusion occurs. And this confidence is used among occlusion detection, to overcome the problem of inaccurate occlusion determination when blocked by analogue. The experimental results show that the proposed algorithm is more robust in the case of the cover, but also has good performance under other complex scenes.
Occlusion Detection for COVID-19 Identification: A Review
