Locomotion Model-Based ATR for GroundSurveillance Radar

摘要

A new classification approach for ATRin ground surveillance radar, based on target locomotionmodeling is proposed. The focus is on a binaryclassification problem: walking person and vehicle targets.The human locomotion model is developed usingbiomechanical model of the human body. The vehiclelocomotion model is simulated for constant acceleratedmotions. A synthetic database was generated using thedeveloped target locomotion models. The database foreach target class was parameterized by cepstral coe?-cients and their distribution was estimated via GMM.Using real data records a classification rate of 94% wasobtained by the locomotion model-based classifier.