Near-online multi-target tracking using aggregate local flow descriptor (ALFD)

摘要

Capturing a video sequence and detecting data coupling between detection and target, where detection is generated using one or more image-based detectors (tracking by detection) Identifying an important target or targets and predicting individual motion, applying an aggregate local flow descriptor to accurately measure the similarity between a pair of detections, and A system and method for tracking a target in a video by applying near online multi-target tracking to perform multi-target tracking given a video sequence is disclosed.