Kinect骨架准确度 [英] Kinect Skeleton Accuracy

问题描述

你好，

我试图用深度图叠加Kinect骨架并意识到这个位置每个关节都非常不准确。

我使用了以下设置：

平滑= 1.0f，

更正= 0.1f，

               预测= 0.1f，

                JitterRadius = 0.05f，

                MaxDeviationRadius = 0.05f

我不认为我犯了编程错误。

骨架在其参照系中有一个非常逼真的行为。  ;但是，我觉得Kinect SDK无法提供与深度图匹配的准确关节位置。

您能否确认或告知我错在哪里？

谢谢！

解决方案

关于深度的9篇文章可能有用的数据：

http ：//channel9.msdn.com/Series/KinectSDKQuickstarts/Working-with-Depthurismo

此外，频道9确实更新了他们完成的骨架项目的样本网站，他们可能会帮助你更多。

另外，确保你从kinect设备至少5英尺左右站立，以获得正确的读数和关节/人类识别。

Hello,

I have tried to superpose the Kinect skeleton with the depth map and realized that the position of each joint is quite inaccurate.

I have used the following settings :

Smoothing = 1.0f,

Correction = 0.1f,

                Prediction = 0.1f,

                JitterRadius = 0.05f,

                MaxDeviationRadius = 0.05f

I do not think I made a programming mistake.

The skeleton has a very realistic behaviour in its frame of reference. However, I have the feeling that Kinect SDK cannot provide accurate joint positions that match the depth map.

Could you please confirm this or tell me where I got wrong?

Thanks!

解决方案

Heres a channel 9 article on depth data that might help:

http://channel9.msdn.com/Series/KinectSDKQuickstarts/Working-with-Depth-Data

Also, channel 9 does have updated samples for the skeleton projects they done on the website and they might help you even more.

Also, make sure that your standing back at least 5 feet or so from the kinect device to get correct readings and joint/human recogition.

这篇关于Kinect骨架准确度的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！