基于机器视觉的弹药零件动态抓取技术研究

Study on Vision-Based Dynamic Grasping Technology for Ammunition Components

针对传统军工业生产制造中,机器人在动态抓取过程中出现的检测精度不高、位姿适应能力差的问题,提出了一种基于深度学习的动态抓取技术,以在传送带运动过程中实现对弹药零件的高效抓取。采用YOLOv5目标检测网络,以实时准确地识别和定位弹药零件,结合卡尔曼滤波算法进行精确的目标跟踪和位置预测。此外,通过实施逼近抓取策略,确保在动态环境中实现稳定可靠的抓取。以机器人操作系统(ROS)为平台进行方案可行性验证。结果表明,该动态抓取方案可满足实际需求,为类似抓取任务提供了参考。

In response to the issues of low detection accuracy and poor pose adaptability in the dynamic grasping process of robots in traditional military industry manufacturing,this paper proposes a dynamic grasping technology based on deep learning to efficiently grasp ammunition components on a moving conveyor belt.The YOLOv5 object detection network is employed to accurately and in real-time identify and locate ammunition components,while the Kalman filter algorithm is utilized for precise target tracking and position prediction.Additionally,an approaching grasp strategy is implemented to ensure stable and reliable grasping in dynamic environments.The feasibility of the proposed scheme is verified using the robot operating system(ROS)platform.The results indicate that this dynamic grasping solution meets practical requirements and provides a reference for similar grasping tasks.