Multi-object tracking

• Main Authors: PANG SU, GERARDO CASTRO MARCOS PAUL, JINHYOUNG OH, JAIN, JINESH, THOMAS, CLIFTON, K
• Format: Patent
• Language: Chinese; English
• Published: 05.09.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The present disclosure provides "multi-object tracking". At a first time step, one or more first objects in the first fused image may be determined based on determining one or more first radar clusters in the first radar data and determining one or more first two-dimensional bounding boxes in the first camera data. A first detected object and a first undetected object may be determined by inputting the first object and the first radar cluster into a data association algorithm, the data association algorithm determining a first probability, and adding the first radar cluster and the first object to one or more of a first detected object or a first undetected object by determining a cost function. The first detected object and the first undetected object may be input to a first Poisson-Bernoulli Hybrid (PMBM) filter to determine a second detected object, a second undetected object, and a second probability. The second detected object and the second undetected object may be reduced based on the second probabilit