HALCON12: Generating 3D Surface Models for Surface-Based 3D Matching
HALCON’s surface-based 3D matching method allows users to locate an object in a 3D point cloud scene and determine its pose by using a trained 3D surface model of the object as the search template. The surface model can be generated from a CAD drawing of the desired object or from a 3D point cloud of the object itself.
HALCON supports several methods for reconstructing a 3D scene including binocular stereo, multi-view stereo (more than 2 cameras), time-of-flight, laser triangulation (sheet-of-light), structured light, and more. Any of these imaging methods can be used to capture the 3D data required for generating a 3D surface model for HALCON’s surface-based 3D matching.
The first step in creating a 3D surface model is capturing a series of 3D images of the entire object. Figure 1 below shows some of the series of images taken of an object as it is being rotated beneath the 3D sensor.
By matching and aligning the overlapping portions of the 3D sensor data, the relationship between the coordinate systems of each object model is determined and can be transformed into a single reference coordinate system. This process is called registration. Figure 2 below shows 2 images and the corresponding registration and transformation of the associated 3D object models. Using a large series of images, enough points can be collected and registered to build up a detailed 3D surface profile of the object. The registered object models are then combined into a single 3D object model.
As shown in Figure 3 below, the unified 3D object model is then sub sampled and filtered (smoothed) to remove outliers and even out the point density of the model. It is then triangulated to create a solid 3D surface.
Next, the object is segmented from the background based on characteristic features of the object, e.g. volume, x,y,z position, etc. (Figure 4) This provides us with the isolated 3D surface model of the desired object which is then trained as the search model for surface-based 3D matching. (Figure 5)
Interested in HALCON 12? Find out more and download the Data Sheet
For detailed information on generating 3D surface models, read section 2.3 in the HALCON Solution Guide III-c 3D Vision.