Vision-based object recognition plays an essential role in analyzing and understanding information obtained from vision sensors for advanced driver-assistance systems and autonomous mobile robots. To realize practical object recognition on embedded systems is not an easy task because these systems need to satisfy high recognition accuracy, real-time processing, and low power consumption simultaneously. For developing this type of embedded systems, hardware-oriented algorithm developments, software-based verifications, and high-performance hardware accelerations are indispensable. In our lab, we are studying on a broad range of research fields including these tree components and realizing working systems for practical applications.

【Software-based Pedestrian Detection & Tracking】

【Hardware-based Realtime Object Detection】