This paper discusses image processing algorithms for the recognition of environmental and road conditions from real-time camera images. This research addresses various implementation techniques for and considerations of the implementation of algorithms to extrapolate various features from images taken by stationary traffic cameras. These algorithms have uses in the areas of traffic pattern analysis, emergency response, real-time traffic situation awareness, and homeland security. This paper covers the design and implementation of an automated camera heading detection system to determine the directional components of a camera's position using the current camera image, various computer vision techniques, and a series of classification training images. Image processing topics addressed as part of this research include edge detection, line detection, and two-dimensional filtering. This paper discusses the implementation of a vanishing point based camera heading detection system capable of analyzing the two dimensional heading from traffic camera images. Although this implementation performed well for the problem of detecting two dimensional vanishing points and camera rotational heading in a 2D image plane, it must be further adapted to detect 3D characteristics and account for the image acquisition distortion common to video monitoring systems. In addition, the Hough Transform implementation discussed suffers from high computational complexity which may hinder use in a realtime video situation. Adaptation to realtime camera video would likely benefit from the implementation of optical flow methods (Shapiro and Stockman 2001) across multiple image frames in addition to the methods described in this paper. The perspective projection vanishing point determination technique discussed within this paper allows for simple vanishing point detection and is easily adapted to the detection of multiple vanishing points. Additionally, the secondary Hough Transform method allows for clear distinction of multiple sets of parallel lines without the difficulties common to simple edge detection techniques. Future study on this topic will analyze the 3D camera position and consider the calibration information for the individual traffic camera position. In addition, this research will be adapted to analyze images from curved road positions. It is expected that the multiple instances of parallel line segments present in curved road sections will be easily detectable within the two HT parameter spaces.


    Zugriff

    Zugriff über TIB

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Realtime image processing algorithms for the detection of road and environmental conditions


    Beteiligte:
    Haran, James G. (Autor:in) / Dillenburg, John (Autor:in) / Nelson, Peter (Autor:in)


    Erscheinungsdatum :

    2006


    Format / Umfang :

    6 Seiten, 7 Bilder, 12 Quellen



    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Print


    Sprache :

    Englisch




    Realtime Image Processing Algorithms for the Detection of Road and Environmental Conditions

    Haran, James G. / Dillenburg, John / Nelson, Peter | ASCE | 2006


    Realtime Image Processing Algorithms for the Detection of Road and Environmental Conditions

    Haran, J. G. / Dillenburg, J. / Nelson, P. et al. | British Library Conference Proceedings | 2005


    Implementing Image Processing Algorithms on FPGA-based Realtime Vision System

    Hirai, S. / Zakouji, M. / Tsuboi, T. | British Library Conference Proceedings | 2003



    Realtime On-Road Vehicle Detection Approach Based on Cascaded Structure

    Zhang, Liang Xiu ;Qiu, Xu Yun ;Zhang, Zhu Lin | Trans Tech Publications | 2011