摘要
抓取检测模块、抓取规划模块和控制模块组成了机器人抓取系统。其中,抓取检测模块是整个系统的重点部分。抓取检测的精确,决定后续的抓取是否成功。本文给出一种基于YOLOv3的端到端的抓取检测方法,该方法改进了特征提取网络和损失函数,建立Grasp-YOLO抓取检测网络,在所建立的样本数据集上进行训练,并在样本测试集上对算法进行验证。实验结果表明,基于YOLOv3的抓取检测算法可以有效地提高检测的精度和速度。
The robot grasping system is composed of grasping detection module,grasping planning module and control module.Among them,the grasping detection module is the key part of the whole system.The accuracy of the grasping detection determines whether the subsequent grasping is successful.This paper presents an end-to-end grab detection method based on YOLOv3.This method improves the feature extraction network and loss function,establishes a grasping detection network called Grasp-YOLO,trains on the established dataset,and verifies the algorithm on the sample test set.The experimental results show that the grasping detection algorithm based on YOLOv3 can effectively improve the detection accuracy and speed.
作者
朱江
李华健
Zhu Jiang;Li Huajian(The 28th Research Institute of China Electronics Technology Group Corporation,Nanjing 210007,China)
出处
《信息化研究》
2022年第1期19-24,共6页
INFORMATIZATION RESEARCH
