The training process in the Teachbench programme is very easy as there is no need for real world learning samples, because the CAD file contains all information. The training supports the user for selecting the range of possible poses and detects „symmetry collisons“ during training.
CAD-based 3D object recognition tool
CVB DNC allows to locate 3D objects described by a CAD file in point clouds and determines the positions of the objects found. Only the object's geometric properties are taken into account, its colour and texture do not matter. This is where the tool's name comes from: Depth No Color.
DNC is a two-stage detection tool.
In a first step, the CAD object to be found is trained. This teaches how the object will later be visible to the sensor.
After the learning process, the object is detected in point clouds in a second step.
Object detection results in a list of hits, each indicating the object's location and orientation within the point cloud. If the sensor generating the point cloud is calibrated with a robot, the hit data can be transferred to the robot in order to grip the objects.
Reasons to choose CVB DNC
The detection of objects is very fast. The detection results can be directly used for robot navigation.
The DNC tool is available for Windows and Linux (including ARM platforms). This makes CVB DNC available in embedded applications.