Phase I Proposal #B045-014-0221: Target Scene Resolution and Calibration Using Innovative Region Based Active Contour Techniques

Given the stressing timelines of KEI, there is a need to quickly and accurately determine the precise location of the physical centroid of a ballistic missile hard body in order to find the desired point of intercept for the kill vehicle. During terminal homing, when the interceptor relies on its IR seeker, there are many events within the seeker field of view (FOV) that can contribute to a failure to intercept. Some major events and associated IR image phenomena are: (1) Booster intensity overwhelming the intensity of the hard body, (2) re-entry vehicle (RV) separation from a post boost vehicle (PBV), and (3) transient shock wave for an endo-atmospheric intercept. Our algorithms differentiate several objects of different intensities, using the image data directly to find the centroid of the hard body. In this context, the only detection heuristic is that there may be more than one object within the images. The proposed image processing algorithms will be combined in Phase II in an integrated fashion with the physical dynamics of the plume, shock wave, and hard body. Our algorithms include feature detection, identification, and tracking. Approaches that we investigated in the Phase I include the following: (1) image segmentation (active contour methods) to detect the separate parts of the boosting missile, (2) use of the object region rather than edge information for tracking, (3) consideration of image noise, and (4) use of shapes for further discrimination and classification. This approach has been successfully demonstrated in a laboratory environment on medical MRIs and commercial aviation applications, as well as on data generated by missile simulation programs CHARM and CHAMP. We have demonstrated in Phase I that our region based active contour techniques are robust under noisy conditions. We have also demonstrated that our algorithms are general and flexible, and can be used for single or multi-color seekers and different bandwidths with minimal parameter modifications. For a multi-color case, each color can be processed independently or in parallel, and the images fused subsequently. Or, the multi-color image can be processed simultaneously to render a single segmented result. In addition to the KEI PTHH problem, the imaging algorithms and prototype software that we developed in Phase I were applied to RV midcourse aim point selection rendering promising results also. The key feature of our approach is that it is a general and robust image recognition/segmentation technique that is applicable to an EO/IR sensor in the BMDS environment. During Phase I of this STTR, we worked closely with the MDA Project Hercules and major DoD contractors, who have endorsed our study for STTR Phase II. We intend to develop core and generic capabilities using innovative active contours techniques during Phase II to help solve the challenging problem of PTHH in boosting targets. Based on promising preliminary results in Phase I, we believe we can tailor our algorithms in Phase II and apply them successfully to the RV aim-point selection problem in mid course and terminal phases of flight, as well as to Target Object Mapping (TOM). Our objective is to develop detailed algorithms and tools for potential insertion into BMDS block upgrades and for use in other domains, focusing on missile/air defense applications in Phase II. We intend to extend our approach to non-DoD government applications and to private sector commercial markets in Phase III, in particular using this technology in the medical diagnostic fields (which the UCLA Medical School has shown great interest in collaboration) such as cancer and cardiovascular disease diagnostics.