Date of Publication :2nd June 2024

Abstract: In the realm of computer vision, detecting objects stands as a crucial task with broad applications, ranging from autonomous driving to surveillance. Acknowledging its pivotal role, we introduce an innovative approach to optimize algorithms for object detection specifically tailored for Field Programmable Gate Arrays (FPGAs). FPGAs are esteemed for their flexibility and ability to process tasks in parallel, making them an ideal platform for speeding up demanding tasks such as object detection. Our optimization strategy is carefully designed to boost the performance of FPGA devices while keeping resource usage to a minimum. We present a detailed methodology that covers architectural considerations, algorithm enhancements, and strategies for hardware-software co- design, all aimed at adapting and refining object detection for FPGA implementation. As part of our approach, we integrate edge detection functionality by utilizing well-known algorithms like the Sobel edge detector to identify edges in images. Additionally, we integrate Harris corner detection to identify significant corners or points of interest, further enhancing object localization capabilities. Through thorough experimentation, we showcase the effectiveness of our approach in achieving real-time object detection with high precision on FPGA hardware. This sets the stage for efficient deployment across various applications, including autonomous driving and surveillance.

Reference :