Multiple Kinect Cameras Skeleton Fusion

Our Goal is combining “Body Skeleton” data from multiple Kinect cameras looking at the same object from different angles.

[wc_button type=”primary” url=”https://www.dropbox.com/s/f3gefhfcsfhh76c/Final%20Submission.rar?dl=0″ title=”Download Project” target=”self” position=”float”]Download Project[/wc_button]

Technical Goal

Our main technical issue was that the Kinect camera does not support in connection of multiple cameras into one computer.

We solved this issue with implementation of a Client software which records one kinect camera on one computer.
The records including 2 time stamps:
the kinect clock and the computer clock.

After that, we implemented a Server software that collect the records from the “Clients” in purpose of merging the skeleton data from each one of the records.

Data combining

Once we overcome the technical problems,

We hold information from each of the cameras synchronised to a “matching frames” list.

Our mission is to combine the skeleton data from each matching into a merged skeleton.

We used “3D Rigid transformation” algorithm wrapped with RANSAC iterations method.

This merged skeleton has better coverage of the body becasue it basically choose the best points in each skeleton frame.

Usages

After gathering all these information from each camera and combining it,

We would like to extract features on this data.

For example, The combined Skeleton data holds in it a motor behavior record.

Hopefully, a more accurate record compared to one kinect camera,

While we will record a person with all cameras, this motor behavior data could be used for diagnosis of motor diseases.

ad7fa9_ffc3385630e24748b11f16a7e4c9e654.png_srz_p_669_532_75_22_0.50_1.20_0.00_png_srz

Client Software illustration

ad7fa9_7a3bcdf1f7654c419eb9e26b7d03ca1c.png_srz_p_629_304_75_22_0.50_1.20_0.00_png_srz

Results – Server application Screenshot

ad7fa9_d4d131b1c6ca40708ee8f6c601dc6020.png_srz_p_911_518_75_22_0.50_1.20_0.00_png_srz

Accessibility
Close Close