How many joints in the body can the RealSense Camera track?

[VPude]

I am thinking along the lines of what the Microsoft Kinect camera can do in tracking certain joints in the body. How does this camera track the body? Is there a set number of joints it tracks? what are those joints? How far does the individual have to be from the camera for it to track them (minimum and maximum distance)? how many bodies can it track at once?

[MartyG · Accepted solution]

Apologies for the delay in responding. I was giving thought to the best way to answer your questions.

1. How does the camera track the body?

This link may be useful to you in explaining that.

https://iq.intel.co.uk/how-the-realsense-3d-camera-works-and-some-clever-things-you-can-do-with-it/ https://iq.intel.co.uk/how-the-realsense-3d-camera-works-and-some-clever-things-you-can-do-with-it/

2. Is there a set number of joints it tracks?

As mentioned in the link above, RealSense can track 22 joints in the hand. It can also track 78 facial landmark points.

This link provides much deeper explanation of hand and face tracking.

https://software.intel.com/en-us/blogs/2015/01/19/face-tracking-using-the-intel-realsense-sdk-bkms Face and Head Tracking using the Intel® RealSense™ SDK | Intel® Software

3. What are those joints?

This link provides the details of the hand joints and face landmarks.

http://www.intel.com/content/www/us/en/support/emerging-technologies/000006099.html Recognition Capabilities of the Intel® RealSense™ Camera

4. How far does the individual have to be from the camera?

The SR300 camera has a min-max depth sensing range of 20 cm to 1.5 m. The R200's min-max range is 30 cm to approximately 4 m. If you want to do hand tracking, the SR300 is the best choice of camera.

5. How many bodies can it track at once?

RealSense can follow one pair of hands (left and right hand), You can track up to four faces with marked rectangles for face boundaries, and select a particular face to focus on. Only one of those faces at a time will have detectable tracking landmarks though.

[VPude]

I think I found a link that is more along the lines of what I am looking for:

https://software.intel.com/sites/landingpage/realsense/camera-sdk/v1.1/documentation/html/index.html?skeletonpoint_personjoints_pxcpersontrackingdata.html https://software.intel.com/sites/landingpage/realsense/camera-sdk/v1.1/documentation/html/index.html?skeletonpoint_personjoints_pxcpersontrackingdata.html

is there a good demo or more documentation I could look at on full body movement? Also any documentation on the limitations or issues? for instance how does the camera handle when a joint moves behind another joint? How is the body tracking when you turn completely away from the camera?

[MartyG]

When a joint is obscured or you turn completely away from the camera, the tracking stalls as the camera loses sight of the tracked joint.

My own project is a good example of full body movement.

https://www.youtube.com/watch?v=T_V8SNwLTrY 'My Father's Face' Tech Trailer 7 - YouTube

[MaryT_Intel]

Hi VPudelski,

Was MartyG's information what you needed or do you still need support from Intel? Let me know and I can hunt down a demo or whatever additional documentation you may be looking for. Thanks,

Mary T.

Support Community Manager