Xpatch Traceback Modification for Improved Signature Estimation and Model Validation

The use of synthetic data for the generation of target signature templates is a vital element of current and future automatic target recognition systems. [1] [2] Although current synthetic signatures closely match their measured counterparts, they suffer from two shortcomings: 1) susceptibility to model error and 2) poor representation of signature variance. The current model validation relies on visual comparisons between the target model and pictures of the target. This technique could be greatly improved if individual facets, which were primary contributors to the observed range profile error could be identified. Such a technique would automate many aspects of the model validation process. The second shortcoming of the current synthetic signature generation is its poor representation of the nondeterministic characteristics evidenced in measured data. These effects are crucial in defining accurate variance estimates of the observed measured data. In order to better represent such nondeterministic effects, ray history information can be used to predict signal variance and increase the accuracy of the classification algorithm. However, the computational complexity and storage requirements of the current ray history traceback method make its use prohibitive. We propose several techniques to overcome these drawbacks and improve the current synthetic signature estimation and model validation procedures.