Rosella – high precision 3D

Rosella is a software toolkit for high precision 3D scanning, point cloud fusion and reconstruction. Rosella interoperates with ROS2 and enables new systems for industrial and scientific applications to be implemented rapidly.

Rosella delivers full workflows of robotic control, 3D data acquisition and point cloud processing that directly enable high precision 3D scanning. Rosella also contains a set of reconfigurable and extendable modules from which new robotics-integrated computer vision systems can be built. Rosella embeds some of the latest global developments in 3D data acquisition, transformation and manipulation.

• Rosella supports acquiring and optimally fusing multiple point cloud views, enabling a precise and accurate 3D understanding of objects from all angles.
• Rosella incorporates algorithmic optimizers and robotics path planners for the automation of system calibration and 3D acquisition.
• Generalized calibration, de-biasing and de-noising procedures in Rosella enable high-precision data to be extracted from a range of proprietary 3D cameras.
• Rosella delivers surface meshing and surface fitting for point clouds, and CAD comparison procedures for applications in additive manufacturing and industrial quality assurance.

Rosella is more just than a powerful computer vision library — it is a toolkit targeting computer vision for robotics systems and robotics-enabled computer vision.

ROS2 integration

ROS2 is a global open-source software initiative targeting robotic systems operation and control. Robotics, real-time sensing, machine vision, and real‑time AI, can all be integrated and synchronized using ROS2.

Rosella is fully integrated with ROS2. Rosella is therefore out-of-the-box ready to be deployed into robotic systems. Robotics system simulation is also provided by Rosella, via interoperability with the ROS2 Gazebo simulator. This allows for the in-silico design and testing of robotic systems that use Rosella. Rosella modules are also engineered with APIs that are not dependent on ROS2. Rosella in its non-ROS2 variant can be better suited to the needs of computer vision algorithm developers and data scientists.

Key features

Rosella’s integrated elements support:

• Interfacing and control of robot arms and 3D cameras.
• 3D point cloud acquisition.
• Multi-point cloud fusion, delivering whole-of-object point clouds with explicit uncertainty estimates.
• Calibration and point cloud fusion that exploits device and system intrinsics using factor graph approaches.
• Meshing, and smooth surface fitting.
• Point cloud to CAD comparison, and infill volume estimation for additive manufacturing.

New features are added to Rosella based on ongoing science and the needs of emerging applications.

Rosella use cases

Existing applications of Rosella include:

• Scientific 3D scanning, targeting insects, birds and other collections.
• Laboratory robot control, for accelerating new polymers discovery.
• Machinery component repair using robotic additive manufacturing.
• Additive manufacturing quality assurance, comparing as-built to as-designed parts.
• Medical applications, in the diagnosis and treatment of wounds and fractures

Licensing of Rosella is permissive and flexible. We are committed to provisioning as much of Rosella as we can under open-source licences, and to fostering open collaborative work with our peers.

Be one of the first outside CSIRO to try Rosella: read about our Rosella early bird opportunity!