We present a statistical pattern recognition scheme for detecting vehicles
in still images. The methodology involves pattern classification using high
er-order statistics (HOS) in a clustering framework. The proposed method ap
proximately models the unknown distribution of the image patterns of vehicl
es by learning HOS information about the vehicle class from sample images.
Given a test image, statistical information about the background is learned
"on the fly." An HOS-based decision measure derived from a series expansio
n of the multivariate probability density function in terms of the Gaussian
function and Hermite polynomials is used to classify test patterns as vehi
cles or otherwise. Experimental results on real images with cluttered backg
round are given to demonstrate the performance of the proposed method. When
tested on real aerial images, the method gives good results, even for comp
licated scenes. The detection rate is found to be quite good, while the fal
se alarms are very few. The method can serve as an important step toward bu
ilding an automated traffic monitoring system. (C) 2001 Optical Society of
America.