THE main social challenge of automotive industry is to develop low cost advanced driver assistance systems (ADAS) able to increase traffic safety. Since vision is the most used human sense for driving, some ADAS features rely on camera based systems. For instance, lane departure warning and lateral control could be reached by detecting the lane markings of the road using computer vision techniques. Besides, knowing where our lane is can be a useful information to help other processing modules also based on computer vision as vehicle detection for automatic cruise control (ACC). Accordingly, in this paper we address the detection of road lane markings from a forward-facing camera. In fact, many works have been done about, which is not surprising since it is a difficult problem due to different common situations: shadows, vehicles occluding the marks, dirty, ego-vehicle movement, etc. Basically, the different proposed algorithms have a first step to collect evidences of where the lane markings are, and a second step that uses them to fit a lane model. Tracking is also added to minimize disturbances from image clutter and facilitate real-time. Since, ideally, lane markings are white lines over a grey pavement, the first step is usually based on image edges, defined as extrema of the gradient magnitude along the gradient direction. Depending on the posterior processing, it is also possible to work directly with the gradient magnitude as edgeness measure. In all cases, gradient direction can be used to remove edges/edgeness having an orientation outside the expected range of values. However, the gradient magnitude can be also high due to the contrast between the asphalt and road elements (e.g. vehicles) and it can also be low because shadows, wear marks, etc. Moreover, the gradient orientation tends to be noisy because its very local nature. In fact, these usual circumstances are the challenging ones since for a road with low traffic, well painted lane markings, without shadows, etc., almost any minimally well-designed computer vision algorithm can succeed in detecting the marks. For this reason, in this paper the authors propose to explore another different low-level image descriptor. Namely, the so-called ridgeness, which would substitute gradient magnitude with the aim of having a more stable response under adverse circumstances. Besides, it also comes with an associated orientation which is less noisy than the gradient one. Even though this is paid with more computation, real-time is still achieved. In order to show how this low-level descriptor is useful for the task at hand, the authors present ridgeness response in adverse circumstances as well as an algorithm that is able to delimit our lane from the ridgeness information in real-time and detect lane change maneuvers.
Ridgeness for detecting lane markings
Rauigkeitsmerkmale zur Entdeckung von Fahrspurmarkierungen
2005
5 Seiten, 5 Bilder, 5 Quellen
Conference paper
English
Ridgeness for Detecting Lane Markings
British Library Conference Proceedings | 2005
|Evaluation of Shared Lane Markings
NTIS | 2010
|Camera calibration from road lane markings
Tema Archive | 2003
|Lane markings-based relocalization on highway
IEEE | 2019
|