Analysis of traffic environment has recently been an important topic as the interest in traffic safety increases. In the field of automobiles, research and development have been conducted on the forward surveillance system using supersonic waves, laser radars, infrared rays etc. In case of railroad security vision-based techniques are thought to be more effective as they can detect, track, classify and identify obstacles. An observation system had been already developed by ultra telephoto camera images to monitor the front view of train to prevent collision with obstacles. The aim of the present paper is to introduce several versions of the algorithm that extracts rails in the obtained image by the well-known Hough transformation, detects and compensates for camera blurs by analyzing the optical flow, and detects moving or stationary obstacles by using image processing techniques. Optical flows at the same magnitude of yaw and pitch directions were observed with almost all pixels in the image when a camera blur occurred. It was demonstrated that it is possible to compensate for such camera blurs effectively by subtracting these flows. A detection algorithm for moving obstacles was proposed. This algorithm is based on the analysis of optical flow which characteristically occurs in those obstacles as a lump. A detection algorithm for stationary obstacles was developed and the obstacle edge and the brightness contrast difference from the background was focused. Experiments under a variety of rail images were successfully performed. As a result, a 600m distant obstacle could be detected with ultra telephoto camera images. Furthermore, in order to cope with all weather conditions, an experiment using a night view image taken by a low luminance CCD camera was performed to prove the possibility of image processing. One of the most important goals in obstacle detection by image processing is to improve the accuracy of recognition. There are some promising methods, such as those based on the distance information obtained by a stereo camera and color histogram information etc. Further investigations, therefore, have to clarify whether they are applicable to railways. However it is difficult to construct a practical detection system only by using the machine vision method. Therefore, it is important to integrate several sensors that complement different weak points. Offline processing has been performed because it is in a research stage yet. Shortening the processing time is also an important problem to be solved. The optical flow algorithm is normally complicated and requires a long calculation time. Therefore, this method is not suitable for software-based real time processing. The application of exclusive processors to improve the processing time is currently investigated.


    Zugriff

    Zugriff über TIB

    Verfügbarkeit in meiner Bibliothek prüfen


    Exportieren, teilen und zitieren



    Titel :

    Obstacle detection with a sequence of ultra telephoto camera images


    Beteiligte:
    Ukai, M. (Autor:in)


    Erscheinungsdatum :

    2003


    Format / Umfang :

    9 Seiten, 17 Bilder, 2 Tabellen, 2 Quellen


    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Datenträger


    Sprache :

    Englisch




    Obstacle detection with a sequence of ultra telephoto camera images

    Ukai, M. | British Library Conference Proceedings | 2004


    Calibrating Pan-Tilt Cameras with Telephoto Lenses

    Huang, Xinyu / Gao, Jizhou / Yang, Ruigang | Springer Verlag | 2007


    Crossing Obstacle Detection with a Vehicle-Mounted Camera

    Sato, I. / Yamano, C. / Yanagawa, H. et al. | British Library Conference Proceedings | 2011


    Camera Calibration for Lane and Obstacle Detection

    Ernst, S. / Stiller, C. / Goldbeck, J. et al. | British Library Conference Proceedings | 1999


    Design of a Compact Telephoto Objective for 6U CubeSats

    Yazgan, Betul Konakli / Ekinci, Mustafa / Bingul, Ahmet et al. | IEEE | 2023