You are here

Image Segmentation for Target Attitude using a Priori Knowledge

Description:

TECHNOLOGY AREA(S): Air Platform 

OBJECTIVE: Demonstrate contour, shape, optical-flow, or other image segmentation techniques for robust model based test and evaluation of target attitude determination, using perfect a priori knowledge of target geometry, for arbitrary cluttered backgrounds. 

DESCRIPTION: Photogrammetric multi-view determination of target attitude (or pose) can be a simple machine vision problem, given sufficiently resolved imagery and benign backgrounds. For ground-based imagery of missiles in flight, however, resolution can be marginal and backgrounds are rarely benign. For cooperative tests, the external geometry of the missile can be known a priori to whatever accuracy is desired. Additionally, the location of the target missile can generally be localized in the image a priori as well. This is not a tracking issue. What is needed is an evaluation algorithm that provides the “best” overlay of the known physical target on top of the measured scene. Marginal resolution, varying lighting conditions, cluttered backgrounds, and poor target contrast contribute to the difficulty of segmenting the image, and determining the best monocular pose for the target missile. A significant portion of the effort should involve selecting and or creating relevant bench mark test data sets for comparison against other state of the art approaches and methods. 

PHASE I: The Phase I effort should leverage industrial and academic advances to develop approaches to achieving the desired overlay at various image resolutions, signal to noise levels, clutter, etc. The method should be demonstrated on a variety of government-supplied synthetic and real datasets, as well as others. 

PHASE II: A successful research effort will produce the following deliverables: 1) An image processing toolkit suitable for inclusion in current government-owned analysis tools, 2) A report detailing extensive verification of the toolkit using benchmark synthetic and real imagery, and 3) A paper accepted to a relevant scientific conference. 

PHASE III: The Phase III program would commercialize the Phase II product for applicability to a wide range of image analysis topics, such as target acquisition, tracking, identification, and general pose estimation. 

REFERENCES: 

1. “Spatially variant mixture model for natural image segmentation,” Can Hu, Wentao Fan, Ji-Xiang Du, Nan Xie, SPIE J. Electronic Imaging 26 (4), 11 July 2017; 2. “Automatic Image Registration Based on Shape Features and Multi-scale Image Segmentation, “ Haigang Sui et al., IEEE 2017 2nd International Conference on Multimedia and Image Processing (ICMIP)C. Das, Naresh & Olver, Kim & Towner, F. (2005). High emissive power MWIR LED array. Solid-State Electronics. 49. 1422-1427. 10.1016/j.sse.2005.06.018.; 3. Interactive image segmentation based on object contour feature image,” Qiang Chen et al., 2010 IEEE International Conference on Image Processing, pp. 3605-3608; 4. Spatio-temporal image segmentation using optical flow and clustering algorithm,” S. Galic et al., IWISPA 2000. Proceedings of the First International Workshop on Image and Signal Processing and Analysis. in conjunction with 22nd International Conference o

KEYWORDS: Image Analysis, Image Segmentation, Pose Estimation, Missile Attitude 

US Flag An Official Website of the United States Government