中图分类号：TP242.6文献标识码：ADOI:10.19339/j.issn.1674-2583.2026.010.006 中文引用格式：支铁山,刘爱华,王淑娟，等. 基于嵌入式视觉的自动捡球机器人控制系统设计［J］.集成电路应用，2026，43（10）：33-39. 英文引用格式：Zhi Tieshan，Liu Aihua，Wang Shujuan, et al. Design of automatic ballpicking robot control system based on embedded vision［J］.Application of IC，2026，43（10）：33-39.
Design of automatic ball-picking robot control system based on embedded vision
1. School of Intelligent Manufacturing Engineering, Guangxi Vocational and Technical College of Natural Resources; 2. School of Intelligent Transportation, Guangxi Vocational and Technical College of Logistics
Abstract： This paper designs and develops an automatic ball-picking robot control system based on embedded vision for the problem of ball target perception and retrieval in unstructured sports venue environments. The system is intended for complex scenes around tennis training grounds and courts. It focuses on several main difficulties,including missed detection of small ball targets,interference from venue textures,and jitter during mobile chassis operation.At the perception level,the study proposes an improved YOLOv8 detection network that incorporates a CBAM channel-spatial dual attention mechanism and a WIoU dynamic focus loss function. Ablation experiments confirmed that each module contributed on its own,and the model reached an mAP@0.5 of 96.5%. At the control level,a kinematic model of a tracked differential chassis was established,while deriving a path tracking control law based on the Pure Pursuit algorithm,and designing a robotic arm grasping strategy based on Position-Based Visual Servoing (PBVS).
