3D Image Processing

General information

In order to perform handlings tasks autonomously in changing everyday surroundings, a service robot has to be able to recognize objects and calculate their position in 6 dimensions (position and orientation). Where mobile service robots have to carry out manipulation and navigation tasks fully autonomously in known and also unknown changing environments, the robot has to record its environment in 3D. Both moving and stationary objects have to be detected systematically and reliably in the robot’s surroundings.

Expertise and sample projects

Fraunhofer IPA has developed a versatile software library that can be used in different ways to automatically generate 3D environment maps, to teach and recognize everyday objects as well as to indentify, localize and track humans. Besides the basic functions for pre-processing / filtering and fusion of sensor data, it also contains the following:

Methods for reconstructing a three-dimensional environment:

• Registration of point clouds and CAD models
• Extraction of features from color and depth data
• Segmentation and reconstruction of surfaces
• Context extraction

Methods for automatically generating and recognizing object models:

• Calculation of stabile 6D feature points on the basis of an object’s texture or surface
• Fusion of several object views in an object model
• Reliable recognition of any modeled object

Methods for identifying classes of previously unseen objects:

• 3D and 2D segmentation
• Calculation of features and attributes (e.g. color or shape)
• Classification of individual segments (= simple objects)
• Classification of segment groups (= compound objects)

Services

As your competent partner, Fraunhofer IPA supports you throughout the entire development of your customized software components for 3D recognition

• System design and selection of suitable sensors and image-processing technologies
• Adaption of existing or development of new algorithms for your specific application
• Integration of technologies into existing systems and applications