State-of-the-art synthetic aperture radar (SAR) systems rely on the calculation of specific statistics produced by transformation of the underlying polarization states. Since a single value is determined, typically without exploiting any sort of averaging, the result may be subject to a large amount of variance.
Some approaches have addressed this problem through sub-aperture processing, thus increasing the number of available images but this reduces the resolution of the imagery available for subsequent processing. Other methods combine the SAR images from each polarization channel to create a single image for use by downstream target detection algorithms. While achieving optimum performance in terms of a specific measurement criterion, they fail to exploit the highly specific co-polarimetric to cross-polarimetric relationship present in symmetric, man-made objects.
Addressing the above concerns, Army researchers have devised a system to reduce the variability of statistics calculated to detect these symmetric, man-made objects in SAR imagery. This is achieved via the incorporation of averaging and the utilization of non-coherent, magnitude data. The novel algorithm increases the target-to-clutter ratio when targets are symmetric while leaving them nearly unchanged when targets are asymmetric.
The polarimetric SAR receiver comprises four input-output receiver-transmitters basis channels for inputting data into a processor. The data is fully polarimetric and includes:
- Horizontal antenna transmitted data which was received by a horizontal receiver antenna data
- Horizontal antenna transmitted data which was received by a vertical receiver antenna data
- Vertical antenna transmitted data which was received by a horizontal receiver antenna data
- Vertical antenna transmitted data which was received by a vertical receiver antenna data
The SAR sensor produces imagery of high enough down-range and cross-range resolution to ensure that one image pixel encompasses the target’s point of left and right symmetry without including contributions from non-target objects. The SAR images – collected simultaneously at different polarizations (bulleted above) – contain information regarding the polarization state.
The inputted data is then focused to produce four co-registered SAR images of a common scene, wherein a specified pixel value represents the radar cross-section of the same patch of ground in each of the four basis polarizations. The four images are then inputted through channels to the polarimetric man-made object detector which calculates a statistic indicating the likelihood that a man-made object is present.
- Operates on a full-aperture SAR image, thereby maintaining the highest possible underlying image resolution
- Utilizes the complex magnitude of each component image pixel, thus enabling the exploitation of spatial averaging for speckle reduction
- Could also be used in imagery produced by other sensors if multiple channels are available, and the measured signals from targets of interest are larger in certain channels while remaining smaller in others
- US patent 9,885,784 available for license
- Potential for collaboration with Army scientists and engineers