Coupling Vanishing Point Tracking with Inertial Navigation to Produce Drift-Free Attitude Estimates in a Structured Environment

摘要

The Air Force’s need for accurate navigation informa-rntion is often met through the combination of inertialrnand satellite navigation systems. However, the increasedrnuse of smaller and smaller unmanned aerialrnsystems opens the possibility of flight in environmentsrnwhere satellite navigation signals are eitherrnsignificantly degraded or unavailable entirely, suchrnas indoors and in dense urban areas. Fortunately, thernintrinsic structure of man-made environments can bernexploited to aid in determining a vehicle’s attitudernwhen satellite signals are unavailable.rnThis research aims to obtain accurate and stable estimatesrnof a vehicle’s attitude by coupling consumergraderninertial and optical sensors. This goal is pursuedrnby first modeling both inertial and optical sensorsrnand then developing a technique for identifyingrnvanishing points in perspective images of a structuredrnenvironment. The inertial and optical processesrnare then coupled to enable each one to aid thernother. The vanishing point measurements are combinedrnwith the inertial data in an extended Kalmanrnfilter to produce overall attitude estimates. This techniquernis experimentally demonstrated in an indoorrncorridor setting using a motion profile designed tornsimulate flight. Through comparison with a tacticalgraderninertial sensor, the combined consumer-graderninertial and optical data are shown to produce an attitudernsolution that is accurate to within 1.5 degreesrnand free of long-term drift. A measurement bias isrnmanifested which degrades the accuracy by up to anotherrn2.5 degrees.