A Real-time Plan-view Method For Following Bending Roads
Xian Chen, E.L. Dagless, Shujun Zhang, B. T. Thomas · 2005
This paper presents a few methods for improving a plan-view algorithm for finding roads in images used in autonomous vehicle navigation. The boundary of coordinate transformation is divided into a number of linking parts to follow a bending road, and a fast algorithm is developed to generate a transforming grid in real-time. It is found that a new method of edge detection carried out on a texture image yielded from the plan- view image shows a better performance than that of the usual edge detection directly on the grey level image. Second-order polynomials are used to approximate the road edges combined with the flexible boundary, improving the capability to follow a bending road.