ABSTRACT
This chapter shows the effectiveness of 3D reconstruction, and the computational complexity is acceptable as will be stated in the experiments. A two-view geometric model is exploited to describe general scenes with respect to a reference plane, which is suitable for road detection and reconstruction problems and is the first time to be used for road reconstruction. The searching region for 3D reconstruction is significantly reduced owing to the two-view geometric model, improving the computational efficiency and the ability to deal with textureless or repetitively textured areas. The computational efficiency of 3D reconstruction is mainly affected by parameter errors of the vision system, not the extrinsic parameters; thus, the measuring accuracy can be improved conveniently by increasing the baseline between the two cameras without increasing computational burdens significantly. The geometric reconstruction approach is extendable to any calibrated perspective two-camera systems as long as the two cameras have corresponding views.
