Road Configuration Estimation Using Dynamic Object Tracking

The patent applies dynamics by transitioning from static object classification to dynamic tracking. The object correlation unit continuously tracks objects across multiple observation time points, using motion patterns to distinguish stationary objects (road edges) from moving objects (vehicles). This dynamic approach resolves the contradiction by maintaining detection accuracy while improving estimation reliability through temporal analysis of object behavior.

The patent implements feedback through the object correlation unit, which uses past observation results to inform current object identification. By correlating objects across multiple time points and using tracking information to validate current detections, the system reduces erroneous detections from noise and multipath effects, thereby improving both detection accuracy and estimation reliability.

The patent applies preliminary action by performing object correlation and tracking before final road configuration estimation. The object correlation unit pre-processes observation data by identifying and tracking objects across multiple time points, establishing their motion patterns in advance. This preliminary classification ensures that only verified stationary objects are used for road edge determination, preventing moving objects from being erroneously identified and improving both position accuracy and classification reliability.

The patent implements continuity of useful action through continuous object tracking across multiple observation time points. Instead of isolated detections, the system maintains continuous tracking of objects, using this continuous information to reliably distinguish stationary from moving objects. This continuous observation and tracking ensures accurate road edge position determination while preventing misclassification of moving objects.

The patent applies the taking out principle by extracting and removing erroneous detection results from the set of all observation results. The object correlation unit identifies and excludes detections that do not correspond to actual objects or are caused by noise and multipath effects. By extracting only valid stationary object positions from the multiple observations, the system maintains high estimation accuracy while eliminating the harmful influence of erroneous detections.

The patent implements parameter changes by transforming the raw observation data into correlated object trajectories. Instead of directly using multiple observation results, the system changes the parameter representation to include temporal correlation and motion patterns. This transformation allows the system to utilize multiple observations for accurate road configuration estimation while filtering out erroneous detections through parameter validation and consistency checks.