Multi-scale FAST feature extraction heterogeneous design based on FPGA

摘要

Mage feature corner detection, is the first step in many visual positioning algorithms, such as tracking, positioning, slam. The fast feature corner point extraction method can greatly improve the real-time performance of positioning, thereby improving the positioning accuracy of objects in high-speed motion scenes, what's more the accelerated implementation of this process is also emerging through hardware system, and it has be valued and used by more and more developers and researchers, However Resource consumption problem and video stream data make this process difficult, So we design an accelerator by FPGA heterogeneous design scheme to fix the problems and The system processes 640×480 resolution images at 50fps with a delay of 44ms.