Standalone Multiple Anomaly Recognition Technique

Threat materials can take many shapes and therefore this software tool must be able to perform anomaly detection instead of specific shape detection. The problem of solely relying on a database of images and a matching algorithm is that objects in the image may be shifted, rotated, or the image quality is too poor for a reliable match. Thus, this proposal focuses on the development of an anomaly recognition algorithm that is able
to achieve low false positive rates even with low quality input images. Our proposed research will develop an
innovative Standalone Multiple Anomaly Recognition Technique (SMART) to determine the presence of contraband in non intrusive inspection (NII) images of trucks and cargo containers. SMART will be able to position in the NII image the location of the potential contraband by using state-of-the-art spectral decomposition analysis techniques to efficiently differentiate the common background of the image against anomalies. These anomalies, in most cases, represent contraband concealed in the cargo.