A real-time approach to stereopsis and lane-finding

Reports new results we have obtained in applying stereo vision algorithms to the problem of autonomous vehicle navigation on highways. The project consists of two parts: lane extraction and obstacle detection. Our lane extraction system is based on a parameterized model for the appearance of the lanes in the images. This model captures the position, orientation and width of the lane as well as the height and inclination of the stereo rig with respect to the road. A robust lane recognition procedure is employed to recover the position of the vehicle within the lane. This scheme is able to recover and track the lane markers in real time (20 Hz) even in the presence of a significant number of spurious lane features. We have developed a new real time stereo system (20 Hz) that has been optimized for use in a highway navigation system. The algorithm first extracts vertical line features from a subset of the scan lines in the left and right images. Correspondences are made between line features in the left and right images by a region-based correlation scheme. The results of this process can then be passed to the grouping and tracking process. The paper also describes the implementation of these algorithms on our network of TMS320C40 DSPs. This computational architecture allows us to handle the demanding computational and I/O requirements of these applications. These algorithms have been tested on video data obtained from a test vehicle that was driven in typical highway traffic. Results from these experiments are presented.

[1]  John F. Canny,et al.  A Computational Approach to Edge Detection , 1986, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[2]  Takeo Kanade,et al.  Vision and Navigation for the Carnegie-Mellon Navlab , 1987 .

[3]  D.J. Kriegman,et al.  Stereo vision and navigation in buildings for mobile robots , 1989, IEEE Trans. Robotics Autom..

[4]  Surender K Kenue,et al.  Lanelok: Detection Of Lane Boundaries And Vehicle Tracking Using Image-Processing Techniques - Part I: Hough-Transform, Region-Tracing And Correlation Algorithms , 1989, Other Conferences.

[5]  Ernst D. Dickmanns,et al.  Recursive 3-D Road and Relative Ego-State Recognition , 1992, IEEE Trans. Pattern Anal. Mach. Intell..

[6]  Ramesh Jain,et al.  A parallel architecture for curvature-based road scene classification , 1992 .

[7]  Charles E. Thorpe,et al.  Representation and recovery of road geometry in YARF , 1992, Proceedings of the Intelligent Vehicles `92 Symposium.

[8]  I. Masaki,et al.  Vision-based vehicle guidance , 1992, Proceedings of the 1992 International Conference on Industrial Electronics, Control, Instrumentation, and Automation.

[9]  Dean A. Pomerleau,et al.  RALPH: rapidly adapting lateral position handler , 1995, Proceedings of the Intelligent Vehicles '95. Symposium.

[10]  Alonzo Kelly,et al.  Obstacle detection for unmanned ground vehicles: a progress report , 1995 .

[11]  Sridhar Lakshmanan,et al.  A deformable-template approach to lane detection , 1995, Proceedings of the Intelligent Vehicles '95. Symposium.

[12]  Jitendra Malik,et al.  New results in stereo-based automatic vehicle guidance , 1995, Proceedings of the Intelligent Vehicles '95. Symposium.