Kinect v2 SDK-将人分解为组件

Question

Quick Kinect v2 SDK question here. I'm trying to split up a human into his/her component parts. Basically I would like to write a test app that has 6 live streams (using the Image control in WPF probably mapping to a Bitmap/Writeable Bitmap). These streams should have the background removed (which I think I should do before trying to do the split).

• Stream 1 -> The person's head (including hair)
• Stream 2 -> Left hand
• Stream 3 -> Right hand
• Stream 4 -> Torso (from neck down to ankles)
• Stream 5 -> Left boot
• Stream 6 -> Right boot

Could anyone suggest a way that I might be able to do this, or even tell me if it is possible? I've been banging my head against a wall for weeks now!

Thanks

Benjamin Biggs

Answer 1

What you want to do is perfectly doable. But you will probably need some practice, and I would like to suggest to start with the samples included within the SDK.

You can choice to develop your code in C# or in C++, if you want to use the Microsoft Kinect SDK v2. But if you prefer to use WPF to build your GUI, probably the best solution is to develop in C# .

You can start by taking a look to this post in which is explained how to remove the background from the RGB image using depth data.

Then, use skeletal information to work on each human body parts. For instance, you can think to get the Head skeletal joint, and center a square on it. Then you can print the resulting head (without the background) in a separate WPF window.

For more information about skeletal joints, refer to the BodyBasics-WPF sample, or take a look to this guide .

You can also have the need to use some image processing library. If it will be your case, C# can use most of the OpenCV facilities via the EmguCV library (which is a porting of OpenCV in C#).

For more information, refer to the official Microsoft Kinect SDK 2.0 documentation .