Design and Implementation of Real-Time Target Recognition System for FPGA-Based Tennis Serving Robot

With the rapid development of artificial intelligence, object detection has become essential in fields like surveillance, autonomous driving, UAVs, and robotics. However, traditional methods often have high requirements for computational performance, limiting their deployment on edge devices. To address this issue, this paper proposes an FPGA-based object detection accelerator, which integrates stream architecture and DSP packing to improve computational efficiency and reduce latency. The detection model is efficiently mapped onto the hardware, significantly reducing the reliance on external memory. Moreover, this accelerator is applied to an intelligent tennisserving robot. The robot uses object detection to dynamically adjust its serving strategy, improving adaptability in various training scenarios. The experimental results show that the robot featured with a dual-friction-wheel serving mechanism and fourwheel differential control with a PID system can precisely regulate serving angle, speed, and spin for efficient and intelligent training.