Research on Motion Control Technology of Distribution High Voltage Operation Robot Based on Deep Learning
With the rapid development of artificial intelligence,somatosensory control has become a hot topic in robot human-computer interaction.How to quickly and accurately recognize human posture is a major challenge in achieving somatosensory control.This study will use the improved YOLOv4 model to detect the human body framework,and the improved stacked hourglass network model to identify joint points,in order to improve the speed and accuracy of human pose recognition;According to the characteristics of the robot's upper arm movement and lower body gait control,a joint point mapping algorithm is developed to control the robot's Pro-gressive aspect sense,so as to solve the problem that the power distribution high-voltage operating robot is prone to fall when interac-ting with people.The results showed that the best result of detecting human targets using the improved YOLOv4 network was 84.37％,the convergence loss function of the improved stacked hourglass network model was 0.096,and the PCK value was 88.3％;The average recognition speed of the research model increased by 21.5 seconds compared to the CPN model,indicating that the re-search model not only improves the accuracy of human pose recognition,but also improves the efficiency of somatosensory control,which has certain research value in the field of somatosensory control.
YOLOv4stacked hourglassjoint point recognitionrobotsomatosensory control
万方数据
维普
