Multilook Fusion using the Sequential Probability Ratio Test
by
Mark W. Koch
Sandia National Laboratories*, Albuquerque NM

With increases in computer processing speed, we can now consider combining multiple video frames to classify a moving object. The sequential probability ratio test (SPRT) allows us to combine multiple classification decisions to decrease the false alarm and missed detections errors. We have developed extensions to the SPRT that allow us to handle dependent observations, video frames with the moving object partially obscured, and an unknown class.

In surveillance, one important problem is to determine whether a moving object is a vehicle or not. Vehicle classification is a challenging problem, since vehicles can take on many different appearances and sizes due to their form and function, and the viewing conditions. The low resolution of uncooled-infrared video and the large variability of naturally occurring environmental conditions can further exacerbate the problem. We have developed a multilook fusion approach using the SPRT for improving the performance of a single look system. Our single look approach is based on extracting a signature consisting of a histogram of gradient orientations from a set of regions covering the moving object. We use the multinomial pattern matching algorithm to match the signature to a database of learned signatures. Using real infrared video and the SPRT we show excellent classification performance, with low expected error rates, when using at least twenty-five looks.

_____

*Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy under Contract DE-AC04-94AL8500.

Date received: March 7, 2007