Implementation of a Hough Transform on a Field Programmable Gate Array

This presentation describes an undergraduate research project to study hardware implementation of artificial intelligence for particle physics tracking. In an effort to move neural network offline computations closer to the readout electronics of the detectors, pipelining and high-level synthesis design optimization are used in a Field Programmable Gate Array (FPGA). FPGAs have been shown to reduce latency and the large number of processing units that neural networks rely so heavily on. In this presentation, we generate random straight-line tracks in a 256x256 pixel array. Our FPGA performs a Hough transform in order to find the parameters of these lines, much like particle track reconstruction, and the timing capabilities of the algorithm are reviewed. The HLS software Vitis was used to determine clock frequencies and resource utilization estimates necessary for running a Hough transform and compare this to non-FPGA and potential neural network implementations.