Automated extraction of queue lengths from airborne imagery

We propose a methodology to obtain intersection queue length estimates from an airborne image by using connected component analysis of the thresholded region of interest. The queue estimates are obtained by two techniques: by counting individual vehicles in the queue and by the area of a polygon containing the vehicles in the queue. The techniques are tested rigorously on a set of images taken from the cities of Tucson and Seattle and the results are presented.