Research on the Application of FPN Algorithm in Grasping Control of Visual Perception Robot
Aiming at the accuracy control of object grasping by visual perception robot,on the basis of grasping pose estimation,the densely connected feature pyramid network(FPN)is used as the feature extractor to fuse the high-level feature map with stron-ger semantics and the low-level feature map with higher resolution.The grasping process of the robot human body is divided into two stages:the first stage generates the area to be grasped,and the second stage refines the grasping area to predict the grasping pose.The model is trained on Cornell and Jacquard data sets,which verifies the effectiveness of the proposed algorithm in grasp-ing pose estimation.Two kinds of real scene object grasping control experiments are designed.The results show that the proposed model can effectively improve the robot's ability to grasp objects of different sizes.
NETL
NSTL
万方数据
维普
