One of the challenges facing the home health care use of robotics is equipping the robot with the ability to differentiate its assigned patients from the rest of the environment, whether that be other family members or the family dog.
At the annual meeting of the Minimally Invasive Robotic Association in San Diego, James Ballantyne, a research assistant in the division of surgery, oncology, reproductive biology and anesthetics at Imperial College, London, discussed a computer-based technique being studied for distinguishing people in the environment from surrounding objects. It uses a time-of-flight camera, which provides real-time three-dimensional depth maps of the environment and analyzes the data in three stages: segmentation, shape descriptor construction, and classification.
To date, Mr. Ballantyne and his associates have applied the technology to 144 3-D objects representing people found in everyday activities in an indoor environment. The system correctly identified 91% of the people.
“The system showed high accuracy in correctly identifying people in complex environments in the testing set,” the researchers noted in their abstract. “Future work will aim to fully test the system in everyday activities as the robot navigates around the environment. Furthermore, the system will be enhanced to enable identification of specific patients who need monitoring.”
Now that’s some cool technology.
— Doug Brunk (on Twitter @dougbrunk)