Vehicle Sensor Calibration Using Detection Alignment

Instead of determining ground truth based on minimum distance (which requires passing the target), the patent inverts the approach by using maximum distance observations. The calibration is performed by observing the target at the farthest point in the sensor's field of view, which is always accessible for forward-looking sensors regardless of their narrow field of view.

The patent performs preliminary alignment of the vehicle's movement direction with the target object before calibration. By ensuring the vehicle moves directly toward or away from the target, the system establishes a known geometric relationship that enables accurate calibration without requiring the vehicle to pass the target or achieve minimum distance observations.

The patent extracts and eliminates the dependency on vehicle velocity measurements from the calibration process. By using purely geometric relationships based on sensor observations of a stationary target and vehicle position changes, the method removes the error-prone velocity measurement component while maintaining calibration effectiveness.

The patent introduces the stationary target as an intermediary reference object that mediates the calibration process. Instead of directly using vehicle velocity, the system uses the target's known stationary position and the vehicle's observed changes in range and bearing to the target as intermediate measurements that lead to accurate calibration without requiring direct velocity input.