Multilane-road target tracking using radar and image sensors

Tracking of a maneuvering target moving on a multilane road using radar and image-sensor-based measurements is studied. A novel 2-D road representation of an on-road moving target is introduced. A natural description of the target longitudinal and lateral maneuvering behavior in the 2-D road coordinates is given using multiple models. An improved mean-adaptive acceleration model is utilized to describe the longitudinal maneuver modes of the motion. Three estimators based on IMM are developed that use different schemes for fusion of radar and image-sensor-based measurements: centralized, distributed, and sequential. Simulation results are presented that illustrate the performance of the proposed estimators and demonstrate their improved capability compared with a known 1-D road coordinate (mileage) estimator. Furthermore, a hidden Markov model (HMM) formulation and two algorithms for lane-only tracking using lane observations are proposed. A lane observation model is derived from a basic image sensor providing raw observation data. Simulation results show that the proposed HMM-based lane estimators can achieve good performance for lane tracking when only basic image-sensor-based measurements are available.