ORB Feature extraction and matching in hardware

Abstract

The ORB (Oriented FAST (Features from Accelerated Segment Test) and Rotated BRIEF (Binary Robust Independent Elementary Features)) feature extractor is the state of the art in wide baseline matching with sparse image features for robotic vision. All previous implementations have employed general-purpose computing hardware, such as CPUs and GPUs. This work seeks to investigate the applicability of special-purpose computing hardware, in the form of Field-Programmable Gate Arrays (FPGAs), to the acceleration of this problem. FPGAs offer lower power consumption and higher frame rates than general hardware. A working implementation on an Altera Cyclone II (a low-cost FPGA suitable for development work, and available with a camera and screen interface) is described.