Pattern Recognition and Tracking Using Correlation Filters

Four correlation filters, namely, maximum average correlation height (MACH), extended MACH (EMACH), distance classifier correlation filter (DCCF) and polynomial DCCF (PDCCF) are employed to recognize and track single and multiple identical and/or dissimilar targets. The MACH and EMACH filters are used for detection while DCCF and PDCCF filters are utilized for classification. Three different two-step algorithms are developed, based on the combination of the filters, called MACH-DCCF, MACH-PDCCF and EMACH-PDCCF algorithms. In the first step, the input scene is correlated with all the detection filters (one for each desired target class) and the resulting correlation outputs are combined. Then a predefined number of regions of interest (ROI) are selected from the input scene based on the higher correlation peak values. In the second step, a classification filter is applied to these ROIs to identify target types and reject clutters and backgrounds. Moving target tracking is accomplished by applying this technique independently to all incoming image frames. Simulation results are given for the proposed algorithms using forward looking infrared (FLIR) imagery.