Labelled GM-CBMEMBER filter with adaptive track initiation

Abstract

The initial Gaussian mixture cardinality balanced multi-target multi-Bernoulli (GM-CBMeMBer) filter does not provide track information and assumes that the target birth intensity is known as a priori, but in reality the target may appear anywhere in the detection area. Therefore, the study proposes a labelled GM-CBMeMBer filter with adaptive track initiation. First, the filter introduces track label information and selects the measurements of newborn targets in the prediction process. In addition, it also makes full use of the Doppler information of the airborne Doppler radar. Then, the position estimate and velocity estimate of newborn targets are calculated by using position measurements converted and Doppler measurements, respectively. Further the label information in the prediction process is inherited during the update process, and the target states are updated sequentially with the Doppler measurements after having been updated by using position measurements. Monte–Carlo experiments show that the proposed filter can perform adaptive track initiation effectively, with good tracking performance, and provide target track information well.

Keywords

• Gaussian processes
• Doppler radar
• target tracking
• Monte Carlo methods
• filtering theory
• position measurement
• radar tracking
• Monte–Carlo experiments
• newborn target measurement
• velocity estimate
• position estimate
• airborne Doppler radar
• Doppler information
• track label information
• labelled GM-CBMeMBer filter
• target birth intensity
• multitarget multiBernoulli filter
• initial Gaussian mixture cardinality
• labelled GM-CBMEMBER filter
• target track information
• good tracking performance
• adaptive track initiation
• position measurements
• target states
• prediction process
• Doppler measurements