The detection of lane boundary markings using the modified spoke filter

We propose a new method to detect lane boundary markings from the input image. In this method, pairs of parallel line edges of opposite direction with a given range of distance are detected directly. Using the geometrical constraint of parallelism, it reduces the number of false candidates for lane boundary edges, and simplifies the tracing and fitting processes for line approximation. Moreover, the process of detecting pairs of parallel line edges can be implemented by binary logical operations suitable for real-time processing. We have developed an item of hardware called a “spoke filter board” for high-speed detection of pairs of parallel line edges. It is compatible with the PICIP (pipeline-path configurable image processing system) developed by the authors. The system detects lane boundary markings at video rate