电子科技2025，Vol.38Issue(10)：42-51,10.DOI:10.16180/j.cnki.issn1007-7820.2025.10.006
机械臂抓取无人机视觉检测方法
Visual Detection Method for Robotic Am Gasping of Unmanned Arial Vehicles
摘要
Abstract
The automatic recovery operation of coaxial double-rotor UAV(Unmanned Aerial Vehicle)has high requirements for visual detection accuracy and real-time performance.The existing target detection algorithm has in-sufficient detection accuracy and real-time performance,and can't be directly used in edge computing platform.Based on YOLOX(You Only Look Once-X)detection model,a lightweight visual detection model P-mobilenext-YOLO without anchor frame is constructed in this study.Using lightweight inverse residual module and integrating self-attention mechanism,the feature extraction ability of visual detection is improved and the training parameters are reduced.In the UAV capture experiment,the model detection results are processed by camera and hand-eye calibra-tion,and the target is accurately captured.The experimental results show that the number of parameters is 5.36 MB and the processing speed is 67 frame·s-1,which has high detection accuracy and real-time performance on the edge device.Compared with the mainstream model,the proposed model is lighter and faster,providing an effective target detection solution for the automatic recovery of coaxial dual-rotor UAVs.关键词
深度学习/目标检测/共轴双旋翼无人机/YOLOX/轻量化/机械臂抓取/无锚框/注意力机制Key words
deep learning/object detection/coaxial dual-rotor UAVs/YOLOX/lightweight/robotic arm grasp-ing/anchor-free/attention mechanism分类
信息技术与安全科学引用本文复制引用
顾寅,付东翔..机械臂抓取无人机视觉检测方法[J].电子科技,2025,38(10):42-51,10.基金项目
国家自然科学基金(61703277) （61703277）
上海青年科技英才扬帆计划(17YF1427000)National Natural Science Foundation of China(61703277) （17YF1427000）
Shanghai Sailing Program(17YF1427000) （17YF1427000）
