Color-based models for outdoor machine vision

This study develops models for illumination and surface reflectance for use in outdoor color vision, and in particular for predicting the color of surfaces under outdoor conditions. Existing daylight and reflectance models that have been the basis for much of color research thus far have certain limitations that reduce their applicability to outdoor machine vision imagery. In that context, this work makes three specific contributions: (i) an explanation of why the current standard CIE daylight model cannot be used to predict the color of light incident on surfaces in machine vision images, (ii) a model (table) mapping the color of daylight to a broad range of sky conditions, and (iii) a simplified adaptation of the frequently used Dichromatic Reflectance Model for use with the developed daylight model. A series of experiments measure the accuracy of the daylight and reflectance models by predicting the colors of surfaces in real images. Finally, a series of tests demonstrate the potential use of these methods in outdoor applications such as road-following and obstacle detection.