Vehicle Lane Recognition Using Virtual Roads and Sensor Fusion

The patent combines GPS positioning with sensor-based detection (image recognition sensors and radar sensors) to create a hybrid system. The GPS provides coarse location information while sensors provide fine-grained lane-level detection, compensating for GPS errors and achieving accurate lane estimation.

The patent introduces virtual roads as an intermediary concept between GPS coordinates and actual physical roads. Virtual roads are generated based on GPS position and orientation, serving as a reference framework that bridges the gap between GPS data and real-world lane structures, enabling accurate lane identification despite GPS errors.

The patent introduces virtual roads as an additional dimensional reference framework. Instead of relying solely on sensor detection in the physical dimension, the system creates a virtual dimensional overlay that provides structural context (road geometry, lane configurations) to disambiguate sensor data in complex multi-vehicle and multi-lane scenarios.

The virtual road structure serves multiple functions simultaneously: it provides geometric constraints for lane identification, establishes reference frames for sensor data interpretation, and maintains consistency across different driving scenarios (crowded roads, intermediate lanes, intersections), making the system universally applicable to various complex situations.