Improving X-Ray Imaging for the Transportation Security Administration

Our scientific evidence suggests that the number of items in an image affects a screener’s ability to detect a target item; however, a computer cannot count the items using current image processing technology available in X-ray screening. With expert knowledge in human perception and image processing, Battelle scientists identified 40 potential visual complexity metrics. They calculated these metrics for hundreds of X-ray images containing threat objects, such as guns, knives, and improvised explosive devices (IEDs), for which TSA screener threat identification performance measures existed.

Using multivariate statistics, Battelle identified a set of metrics, which accounted for over 70% of correct and incorrect screener decisions. When the visual complexity metrics are calculated for actual baggage, they correlate highly with the number of items actually in the baggage, thus providing the TSA with information about baggage that could not be obtained previously.