Object recognition for mobile manipulation using Kinect camera

This video shows an excerpt of the scenario “Kate cleans up the table” with a new object recognition using a Kinect. In comparison to our previous video (www.youtube.com this video shows the object recognition using the Kinect rather than a 3D laser scan. Only the Kinect and the SmartSoft component for object recognition were replaced. The component is 100% compatible due to its stable component hull.
Video Rating: 5 / 5

Incoming search terms:

• kinect object recognition
• identifying objects using kinect
• kinect object
• object recognition with kinect
• Probabilistic Fusion of Multiple Algorithms for Object Recognition at Information Level
• robocup@home magdeburg 2012 result ulm

Probabilistic Fusion of Multiple Algorithms for Object Recognition at Information Level

This video shows object recognition for mobile manipulation using a Kinect RGBD camera. The robot cleans up the table by dropping all objects on the table into its basket. The objects are recognized by probabilistic fusion of multiple algorithm results. The object recognition is able to run a simple 3D model matcher, color histogramm, barcode and OCR algorithms. The scenario focuses on the recognition of similar objects, such as different flavours of juice. None of the algorithms alone would have been able to identify the objects reliably enough.
Video Rating: 5 / 5

Mobile Manipulation – Robot “Kate” cleans up the table

This video shows a mobile manipulation scenario of our robot Kate cleaning up a table. The scenario was developed at the Collaborative Center of Applied Research on Service Robotics (ZAFH Service Robotics) at the University of Applied Sciences Ulm in Germany. Variants of the scenario were used at the RoboCup@Home GermanOpens 2011 in Magdeburg. The whole system is developed using the SmartSoft MDSD Toolchain. It is one example of how to build a complex system by composing already existing reusable software components. The orchestration of the components is done with SmartTCL. At runtime a task-tree is dynamically created and modified depending on situation and context. Symbolic task planners, like metric-ff, are used to extend or modify the SmartTCL task-tree at runtime. The object recognition is based on a point cloud captured with a HOKUYO laser ranger mounted on the Katana manipulator. The trajectory planning is done with OpenRAVE which takes the mesh representation of the objects into account.

Fembot Chobits Aiko Maid Cosplay Android Robot

Aiko as Chii Chobits Maid Cosplay. Demostration of Aiko Internal OS. AI Female Android Speech, Reading, Color, Face, Object recognition. Ability to tell Weather. Ability to have a conversation Can understand 13000+ sentences, and has the ability to learn. Ability to solve math Ability to distinguish simple drinks and foods, and objects. Mimic human physical touch.
Video Rating: 4 / 5