A robot 3D vision guidance system developed based on structured light measurement technology and 3D object recognition technology can perform fully free positioning and picking of randomly stacked parts within a large measurement depth range. Compared with traditional 2D vision positioning methods, which can only identify parts at a fixed depth and only obtain positional information of some degrees of freedom of the parts, this system has greater application flexibility and a larger detection range. It can provide an effective automation solution for industrial problems such as machine tool loading and unloading, parts sorting, and palletizing.
Machine Vision 3D Guidance System Framework
3D reconstruction and recognition technology
The self-developed 3D scanner can accurately and quickly acquire point cloud images of a scene. Through 3D recognition algorithms, it is possible to identify and estimate the pose of various target objects in the point cloud image.
3D reconstruction and recognition efficiency
Multi-material recognition effect test
Thanks to robust reconstruction and recognition algorithms, it can reconstruct and recognize parts of different materials stably. Even aluminum materials with high reflectivity and black parts can achieve good reconstruction and recognition results, making it suitable for a wide range of industrial scenarios.
Robot path planning
Obtaining the part's pose information doesn't mean you can immediately pick it up; that's only the first step. To successfully pick up the part, the following steps are required:
The independently developed robot trajectory planning algorithm can easily complete the above tasks, ensuring the stability and reliability of the robot's part picking process.
Quickly switch pick objects
With just four simple operations, you can quickly switch between picking objects without having to make complex tooling or production line adjustments.
