跟踪骨架放下 [英] tracking skeleton laying down

问题描述

我注意到，当脊柱（从"Spine"到"ShoulderCenter"的线条）接近90度时，跟踪骨架的误差会增加。

I have noticed that as the spine (the line from "Spine" to "ShoulderCenter") gets closer to 90 degrees the error in tracking the skeleton increases.

例如：如果我站起来面对kinect并且它开始跟踪我的骨架，那么倾斜离开样本 程序很难跟踪我。或者，如果我在桌子上面对kinect，然后躺在我面对kinect的桌子上，它就不会准确地跟踪我的身体（骨架点到处都是，或者根本没有）。

For example: If I stand up facing the kinect and it begins to track my skeleton, then lean left the sample program has a hard time tracking me. Or if I face the kinect at a table, then lay on my side on the table facing the kinect it will not accurately track my body (skeleton points flying everywhere or nothing at all).

到目前为止，我所能看到的只是特定的联合"点"。 API给了我们，如果当一个人不正直时它无法检测骨架上的点，我想不出跟踪某人放下的方法，这基本上是我的问题：

All I can see so far are specific Joint "points" that the API gives us, and if it can't detect points on the skeleton when a person is not upright, I can't think of a way to track someone laying down, which is basically my question:

有没有办法跟踪某人放下的骨架（同时仍然面对kinect）？

Is there a way to track a skeleton on someone laying down (while still facing the kinect)?

推荐答案

Kinect SDK Beta提供的内置骨架跟踪针对直立且不会像墙壁或桌子表面齐平的人进行了优化。即使在这些情况下，有时也可以提取关节，但是
的准确度肯定不是那么好。

The built-in skeletal tracking provided by the Kinect SDK Beta is optimized for people standing upright and that are not flush against a surface like a wall or table. It is possible that joints can be extracted sometimes even in these scenarios, but the accuracy is definitely not as good.

如果你有一个特定的场景，你需要更高的精度联合跟踪一个与墙壁或桌子齐平的主体，然后您可能需要实现自己的特定于域的骨架跟踪算法，该算法使用kinect SDK提供的深度
流中的数据，通过自己做分割和图像处理。

If you have a specific scenario for which you need higher accuracy of joint tracking of a body that is flush against a wall or table, then you will probably need to implement your own domain-specific skeletal tracking algorithm that uses data from the depth stream provided by kinect SDK, by doing your own segmentation and image processing.

祝你好运！

Eddy

Good luck!
Eddy

这篇关于跟踪骨架放下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！