Method of Detecting Lane Curvature with Driver Eye Gaze Vector

摘要

Lateral control features like Lane Keeping Assist and Lane Centering Control will perform better if the algorithm has access to forward looking road curvature information. The current stale of the art uses machine-learned perception in the front camera to estimate curvature. This method is oAen inaccurate and late. Although there is GPS Map data in high-end vehicles, the GPS signal may not be available in all geographic areas. This new method utilizes the driver's eye gaze vector provided by the Driver Monitoring System or in-vehicle cabin monitor to detect lane curvature. The gaze-vector lane curvature can be used with the fusion of front camera data to improve the estimate of forward-looking lane curvature to assist Lateral Control features.