Data-Driven Extraction of Curved Intersection Lanemarks from Road Traffic Image Sequences
Segmentation of optical flow fields, estimated by spatio-temporally adaptive methods, is - under favourable conditions - reliable enough to track moving vehicles at intersections without using vehicle or road models. Already a single image plane trajectory per lane obtained in this manner offers valuable information about where lane markers should be searched for. Fitting a hyperbola to an image plane trajectory of a vehicle which crosses an intersection thus provides concise geometric hints. These allow to separate images of direction indicators and of stop marks painted onto the road surface from side marks delimiting a lane. Such a ‘lane spine hyperbola’, moreover, facilitates to link side marks even across significant gaps in cluttered areas of a complex intersection. Data-driven extraction of trajectory information thus facilitates to link local spatial descriptions practically across the entire field of view in order to create global spatial descriptions. These results are important since they allow to extract required information from image sequences of traffic scenes without the necessity to obtain a map of the road structure and to make this information (interactively) available to a machine-vision-based traffic surveillance system.
The approach is illustrated for different lanes with markings which are only a few pixels wide and thus difficult to detect reliably without the search area restriction provided by a lane spine hyperbola. So far, the authors did not find comparable results in the literature.
KeywordsCurve Section Curve Intersection Edge Element Spatial Description Vehicle Trajectory
- 3.S. Dance, T. Caelli, and Z.-Q. Liu: “A Conceptual, Hierarchical Approach to Symbolic Dynamic Scene Interpretation”, Pattern Recognition, Vol. 29(11), pp. 1891–1903Google Scholar
- 4.J.H. Fernyhough, A.G. Cohn, and D.C. Hogg: “Generation of Semantic Regions from Image Sequences”, in Proc. Fourth European Conference on Computer Vision (ECCV’96), 15–18 April 1996, Cambridge/UK; B. Buxton and R. Cipolla (Eds.), Lecture Notes in Computer Science LNCS 1065 (Vol. II), pp. 475–484, Springer-Verlag Berlin, Heidelberg, New York 1996.Google Scholar
- 5.J.H. Fernyhough, A.G. Cohn, and D.C. Hogg: “Building Qualitative Event Models Automatically from Visual Input”, Proc. ICCV’98, Bombay/India, pp. 350–355, January 1998.Google Scholar
- 6.R. Gerber and H.-H. Nagel: “(Mis-?)Using DRT for Generation of Natural Language Text from Image Sequences”, in Proc. Fifth European Conference on Computer Vision, 2–6 June 1998, Freiburg/Germany; H. Burkhardt and B. Neumann (Eds.), Lecture Notes in Computer Science LNCS 1407 (Vol. II), pp. 255–270, Springer-Verlag Berlin, Heidelberg, New York 1998.Google Scholar
- 7.W.E.L. Grimson, C. Stauffer, R. Romano, and L. Lee: “Using Adaptive Tracking to Classify and Monitor Activities in a Site”, Proc. CVPR’98, pp. 22–29, June 1998.Google Scholar
- 8.F. Guichard and J. Ph. Tarel: “Curve Finder Combining Perceptual Grouping and a Kalman Like Fitting”, Proc. International Conference on Computer Vision ICCV’99, 20–27 September 1999, Kerkyra (Corfu), Greece, pp. 1003–1008.Google Scholar
- 13.H.-H. Nagel and A. Gehrke: “Spatiotemporally Adaptive Estimation and Segmentation of OF-Fields”, in Proc. Fifth European Conference on Computer Vision ECCV’98), 2–6 June 1998, Freiburg/Germany; H. Burkhardt and B. Neumann (Eds.), Lecture Notes in Computer Science LNCS 1407 (Vol. II), pp. 86–102, Springer-Verlag Berlin, Heidelberg, New York 1998.Google Scholar
- 14.N. Johnson and D. Hogg: “Learning the Distribution of Object Trajectories for Event Recognition”, Proc. BMVC’95, pp. 583–592, 1995.Google Scholar
- 15.M. Mohnhaupt and B. Neumann: “On the Use of Motion Concepts for Top-Down Control in Traffic Scenes”, Proc. ECCV’90, Antibes/France, O. Faugeras (Ed.), LNCS 427, pp. 598–600, April 1990.Google Scholar