Mask-RCNN based object segmentation and distance measurement for Robot grasping
Dong-Kyo Jeong, Hosun Kang, Dong-Eon Kim, Jang-Myung Lee · 2019
This paper proposes a new application that the object can be determined by the optimal model. To extract the target from the clutter background accurately, Mask-RCNN (Mask-Region Convolutional Neural Network) model is utilized for the segmentation process. Meanwhile, target object in front of the camera can be localized with the help of the Mask-RCNN segmentation and the geometric stereo matching method. Experiments show that distance values are calculated efficiently. And then the robot manipulator is performed to grasp the target object effectively.