cvl On-Road Vehicle Tracking Based on Mean Shift Estimation
    Ford Motor Company
    YELLOW STRIPS
     


    Project

     

    Experimental Results

    Various on-road tests have been carried out on American highways, including I94 in Michigan and I80 in Reno. Superior tracking performances are obtained by using combined features of color and shape to represent the model and tracking target..

    The following example shows how shape information of the vehicle helps correct the Mean Shift estimator to locate the proper target while there are competing clutters with similar color to the target presented in the background. The first two columns contain the images showing that the original Mean Shift tracker which uses color information alone is deceived by the similar background. The third column shows the image representations in the combined feature spaces of color and shape. The corresponding  tracking results are shown in the fourth column.

     
    Solo Color Representation Mistracked Image Combined Feature Representation Corrected Tracked Image

    The two video sequences to demonstrate the tracking results on US highways are here:

    July 2003 on I94, Michigan

    Oct 2003 on I80, Nevada



    Overview

    Methodology

    Result

    Future Work

    Contact

     
    Other Links

    University of Nevada

    Ford Motor Company


    15 February 2004  by  Kai She