Automatic Target Recognition

The issues addressed in Maryland's ATR research involve sensor modeling, object modeling, combining detection and recognition, and sensor and operator fusion at signal, symbol, and object levels. This research involves four sensors: FLIR, LADAR, Polarimetric SAR (PSAR), and millimeter wave. These sensors complement each other well for fusion purposes, and unclassified data obtained by these sensors is beginning to become available.

The general approach to ATR being pursued at Maryland involves the study of sensor models; statistical models for characterizing background and clutter; geometric and radiometric models for targets; multisensor integration; and performance characterization.

As an initial stage in research on FLIR ATR, a probe based approach has been developed to recognize objects in a cluttered background using an infrared imager. A probe is a simple mathematical function which operates locally on image gray levels and produces an output that is more directly usable by an algorithm. A directional probe image is calculated by taking the difference in gray levels between pixels a set distance apart in a given direction, centered on the probe image pixel. These probe images contain the information necessary for use by an object recognition algorithm in a readily usable, and mathematically describable, form. A parametric statistical image background model which describes the probe images has been developed. The parameters of the probe image model can be readily estimated for these parameters, together with target signatures obtained from Computer Aided Design (CAD) models, allows the likelihood ratio for a given object pose hypothesis versus the background null hypothesis to be wr

Initial work on SAR ATR has dealt with constant false alarm rate (CFAR) detection of targets in fully polarimetric SAR images. CFAR detection algorithms produce many false targets when applied to single-look, high-resolution, fully polarimetric synthetic aperture radar (SAR) images, due to the presence of speckle. A two stage CFAR detector followed by conditional dilation has been used to detect point and extended targets in polarimetric SAR images. In the first stage, possible targets are detected and false targets due to the speckle are removed by using global statistical parameters. In the second stage, the local statistical parameters are used to detect targets in regions adjacent to targets detected in the first stage. Conditional dilation is then performed to recover target pixels lost in second stage CFAR detection.

The performance of a CFAR detector will be degraded if an incorrect statistical model is adopted and the data are correlated. A goodness-of-fit test is performed to decide the appropriate distribution and the effects of decorrelation of the data are considered. Good experimental results were obtained when the approach was applied to single-look, high-resolution, fully polarimetric SAR images acquired from MIT Lincoln Laboratory.