The video above demonstrates robotic assembly based on monocular camera, where the assembly clearance between shaft and hole is 1mm.
The video above demonstrates pose tracking of mechanical parts, where the red dots show the visible edges of the object.
2. Image Demo
Fig. 1
The Fig.1 shows the pose estimation results in open environment of our method。
Fig. 2
The Fig.2 shows the pose estimation results under the different camera pose.
Fig. 3
The Fig.3 shows the process of robot assembly of our method.
3. Key-inside
Yolo
Convolutional Autoencoder
Edge distance tensor
Nonlinear optimization
Opencv
Pinhole camera model
OpenGL
Rigid Object 6D pose description
Lie groups and Lie algebras
4. Method
Fig. 4. Algorithm overview
Fig. 5. Visualization of the algorithm process
5. Contrast
AAE (ECCV 2018): Sundermeyer M, Marton Z C, Durner M, et al. Implicit 3d orientation learning for 6d object detection from rgb images[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 699-715.
