This paper describes a highly efficient and parallelized HW implementation of multiple modules processing a pattern matching algorithm of a 2D affine model object applied to the detection and tracking of an autonomous vehicle. The pattern matching problem has been adapted to the edge detection of the rear side of a Cycab vehicle modeled by the centroid and scale parameters. Reordering of the pattern and online generation of pixels of interest are the main contributions. The computing time for an image is independent of the number of pattern matching processing modules instantiated in hardware, making feasible to use them in large filters. The maximum operation frequency achieved has reached up to 103.1 MHz with minimal use of resources. An implementation of 200 parallel modules is capable to process 200 frames per second for image sizes of 512x512 pixels while equivalent program in a desktop PC executes only 4 fps