Efficient Ego Lane Detection for Various LaneTypes

In this work, we present an ego lane detector de-signed for the use in automotive vision systems for personallight electric vehicles like electric bicycles, tricycles or scoot-ers. The approach is based on a combination of gradient-based line detection, color-based segmentation and geomet-rical rules, making the ego lane detector fast, but also robustto different scenes, including curves. Qualitative evaluationon over fifty traffic scenes show that the lane detector is ableto find a suitable approximation of the road area with an IoUof 75.71%.