弱光环境下仓库搬运机器人抓取控制方法OACSTPCD

To achieve precise,stable,and repeatable grasping actions,a grasping control method for warehouse handling robot based on stable lightweight network in low light environment is studied.Firstly,for handling low light images in the environment,a stable and lightweight encoding/decoding network is used to extract low light features in the grasping area and perform fusion processing to obtain normal light grasping area features.Sec-ondly,in the deep separation fusion extraction layer,by processing the normal light grasping area features,deep features are reconstructed to recover the lost detail information during feature extraction.Then,reconstruction features are input into the network output layer,and the grasping pose parameters of the warehouse handling robot's gripper are output.Finally,the coordinate transformation relationship between the camera coordinate system and the robot coordinate system is obtained through handling target image obtained by hand-eye calibra-tion.The grasping pose parameters are converted into grasping control variables for the warehouse handling ro-bot,completing the grasping control of the warehouse handling robot.The experiment proves that this method can effectively extract the features of the grasping area of the handling target,and effectively predict the grasp-ing posture of warehouse handling robot,complete the grasping control of warehouse handling robot,and have high grasping accuracy.