Embedded vision on Cyclone V

A working system before an accelerated algorithm

The project investigated a face-detection system on the Altera SoCKit board, which combines an ARM processor subsystem with FPGA fabric. The practical starting point was a complete path from a USB camera to a running computer-vision application.

That path crossed hardware configuration, the Linux kernel, peripheral support, the build toolchain, and OpenCV. A capable board on its own did not supply an application-ready environment.

Integration across the stack

The work used Altera QSys and Quartus to establish the board configuration, then a Linux build prepared with the Linaro toolchain. USB Video Class support was part of the kernel work needed to bring a webcam into the system. OpenCV and a face-detection application completed the software path.

The public archive includes the thesis, development material, and demonstrations of board startup and camera operation. Together they show the engineering around the algorithm: configuring a platform, bringing peripherals up, and making its behavior observable.

The result—and the boundary of the result

The integrated application ran on SoCKit and was compared with an Intel Core i7-2670QM system using a standard Lena image. The thesis reports the SoC implementation as approximately 43% slower in that comparison. It is a result for that setup and workload, rather than a general benchmark of ARM against x86.

FPGA acceleration remained a recommendation for further work. The demonstrated contribution was platform and software integration, with a measured baseline from which acceleration could be investigated.

From the laboratory to the publication record

The repository preserves work from 2013–2014. The related paper, “Integration of OpenCV and Cyclone V Hybrid ARM and FPGA SoC for Face Detection Application,” appears in the proceedings of ISCIT 2023. The project date and publication date describe different stages of the same research history.