2D + Time Object Tracking - Satellites

Our team developed an improved Generalized Labeled Multi-Bernoulli (GLMB) filter for tracking objects in satellite videos, addressing challenges where objects appear unpredictably, and detections often include false positives. Our enhancement allows the filter to estimate better starting positions for new objects by learning from past trajectories.

This project enhances multi-object tracking by making initialization more reliable in complex scenarios. Our adaptive birth process significantly improves tracking accuracy in cluttered and unpredictable environments, such as satellite imagery, pushing the boundaries of object tracking in remote sensing.

Our team developed an improved Generalized Labeled Multi-Bernoulli (GLMB) filter for tracking objects in satellite videos, addressing challenges where objects appear unpredictably, and detections often include false positives. Our enhancement allows the filter to estimate better starting positions for new objects by learning from past trajectories.