Real Time Estimation Network for Rigid Body Posture Based on Synthetic Data
There are the characteristics of scarce data,low robustness in complex scenes,and poor real-time for existing rigid body pose estimation,for this reason,a rigid object pose tracking network based on synthetic data is proposed.Temporal and spatial feature fusion techniques are used to capture temporal and spatial feature information,generating spatiotemporal sensitive feature maps.Residual connectivity is utilized to learn more diverse and abstract high-quality features,improving the accuracy of tracking the target.Data augmentation is performed on scarce data to generate complex synthetic data that conforms to the real physical character-istics,which is used to train the deep learning model and improve the generalization of the model.Seven objects are selected on the YCB-Video dataset for real-time pose tracking experiments,the results show that compared with similar related methods,the pro-posed method is more accurate in estimating the poses of rigid bodies in complex scenarios,and it has an optimal performance in real-time estimation efficiency.
万方数据
维普
