A Practical Robotic Grasping Method by Using 6D Pose Estimation with Protective Correction

Pose estimation is the critical technology in industrial robot. Nowadays, many machine vision-based approaches have applied the technology and achieved excellent results. However, the rapid detection of pose estimation in complex multi-scene environments is still a challenge, due to the interference of multi-angle light and multi-background, it leads to remove the object surface features in the image, resulting in the false robotic grasping. To address these issues, this paper proposes a practical robotic grasping method by using 6D pose estimation with protective correction. In this method, the synthetic data set by self-production is used to train the convolutional neural network (CNN), in which the feature extraction network is improved by residual block and optimize branch network by reducing convolution computation. Meanwhile, in order to prevent grasp collisions cause by misrecognition, we propose corrected grasping pose algorithm (CGP) for protective correction in which the 2D keypoints of the object from semantic segmentation network output is to map to the point cloud and calculate the corresponding centroid. Then, the obtained error between the measured translation of the centroid and the predicted translation is used to determine whether it is within the set threshold.