Computational Vision Modeling for Target Detection
Abstract
The current DOD target acquisition models have two primary deficiencies: they use simplistic representations of the vehicle and background signatures, and a highly simplified description of the human observer. The current signature representation often fails for complex signature configurations and yields inaccurate detectability and marginal pay-off predictions for low signature vehicles. In addition it is not extensible to false alarms and temporal cues, and precludes applications to vehicle design guidance and diagnosis. The current human observer model is simplified to the same degree as the signature rnpresentation and as such does not extend to high fidelity largetlbackground signature representations. In answer to these deficiencies, we have developed the TARDEC Visual Model (TVM) that is based upon emerging academic computational vision models (CVM). Recent advances in CVM have made dramatic improvements in the understanding of early human vision processes. A model of neural receptive fields includes a generic image representation of the spatial processing characteristics for early vision cortical areas. An input image is first divided into its three color opponent components with each axis further decomposed into a set of band pass spatial frequency filters (Gabor or wavelet transform filters) with different center frequencies and orientations. Signal to noise statistics are then calculated on each channel, appropriately aggregated over all channels using signal detection theory to predict probabilities of detection and false alarm.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1994
- Accession Number
- ADA479457
Entities
People
- Gary Witus
- Grant Gerhart
- Thomas Meitzler
Organizations
- Tank-automotive and Armaments Command