情報処理学会第85回全国大会 会期:2023年3月2日~4日 会場:電気通信大学

6C-02
1000FPS高速度撮影による顔検出アプリケーションの開発
○桑原譲二,楳田敏之,小川公則,久保田純,植草道人(フォトロン),三島也守志(ヘミボーラ)
This paper describes the development of high-speed face detection application using 1000 FPS camera with USB3 connection. The detection process is implemented with the OpenCV technology called "cascade classifier" and the algorithm consists of, famous, classical but quite effective "Haar-like feature". We explain the fundamentals of the workflow that takes the maximum advantage of the high-speed camera and high-speed computer vision algorithms. It shows the merit of the high-speed facial detector, which cannot be achieved with a traditional camera and application. In addition, we shows how to achieve a massive improvement in processing performance by leveraging the processing cores of the GPU.