Road prediction using video for integrated driver support

摘要

The use of a video camera for predicting vehicle trajectory is reviewed. The data available from the video camera, and the features which can be used for trajectory prediction and how these relate to the real world, are analysed. To accurately predict the road ahead, a model is used, consisting of a spatial and dynamic module. A recursive least square based estimator is designed based on the road model. The estimator is robust with respect to variations in the horizon, deals with distance and varying measurement noise, and handles the spatial/dynamic structure. The results are presented in two separate parts. The first deals with the estimation and comprises simulated data to illustrate robustness, and video data to show typical road features. The second is the performance analysis. It highlights the response to changes in road parameters and the subsequent correlation with standard road prediction techniques.