HALCON’s surface-based 3D matching is a very effective tool for locating the pose of objects by matching a trained surface model to the 3D surface profiles of objects in a 3D point cloud. For robotic applications, this allows users to determine the proper robot picking pose and position of objects regardless of their orientation to the cameras or sensor, and even if they do not have well-defined edges. Where surface-based 3D matching sometimes runs into problems is when the object does not have curved surfaces, but is made up of mostly planar surfaces; for example, a box.
HALCON’s surface-based 3D matching utilises the normal vectors (among other features) of the surface to properly align the 3D object model with the object surfaces in the point cloud. On planar surfaces, the normal vectors all point in the same direction which may result in poor alignment as shown in the image below.
Figure 1. Inaccurate alignment of object model with planar surfaces
With HALCON 13 this problem has been solved. Surface-based matching in HALCON 13 allows the user to extract 3D edges from the point cloud and use those edges along with the surface normals to provide a more robust alignment.
 Figure 2. 3D edges extracted from point cloud |  Figure 3. With 3D edge support, planar surfaces are found robustly |
|---|
As seen in Figure 4 below, sometimes 3D edge data is not precise enough. In this case, even greater precision can be achieved by combining 3D point cloud edge data with 2D edges extracted from an image of the scene (see figure 5). With HALCON 13, surface-based 3D matching with combined 3D and 2D edge support provides the most robust results for finding the position and pose of objects with planar surfaces.
 Figure 4. Sometimes 3D edges are not precise enough |  Figure 5. Combining 3D and 2D edges is even more robust |
|---|
This level of functionality and performance enables much greater flexibility for applications like robotic palletising and depalletising and can reduce the need for costly fixtures and mechanical alignment systems.