问题描述
当一个人与墙壁接触时，骨骼跟踪会在100％的时间内失败。在我们的应用程序中，我们知道每个框架中墙的位置。我们不能告诉骨架跟踪器丢弃pointcloud的某些部分吗？这将打开一个全新的
范围的应用程序，其中人们也在斜躺/躺下。
When a person makes contact with a wall, the skeleton tracking fails 100% of the time. In our app we know where the wall will be in each frame. Can't we tell the skeleton tracker to discard certain parts of the pointcloud? That would open up a whole new range of applications in which people are reclining/lying down as well.
推荐答案
身体追踪（如v2所述，是基于看到头部和肩膀。您必须面对传感器以便跟踪锁定。这是基于传感器看到地板和头部以及用户的双肩。你必须永远不要假设body
索引0是那个人，你需要迭代数组并检查IsTracked属性。如果运行BodyBasics样本，则会出现任何跟踪的关节。你可能希望让这个人离墙一步，然后一旦锁上，那么
就可以移动了。
Body Tracking (as referred to for v2) is based on seeing the head and both shoulders. You must face the sensor for tracking to lock on. This is based on the sensor seeing the floor and the head and both shoulders of the users. You must never assume body index 0 is the person, you need to iterate the array and check the IsTracked property. If you run the BodyBasics sample are there any tracked joints appearing. You may want to have the person take a step away from the wall and then once locked on then they can move.
我们没有提供一种注入数据的方法管道。具体的api和联合属性提供有关过滤数据的信息。如果关联是推断的或未跟踪的，那么您可以在应用程序中以某种方式对待它。
We do not provide a way of injecting data in to the pipeline.The api's and joint properties specifically, provide information on filtering the data. If a joint is Inferred or not tracked, then you can treat that in some way in the app.
这篇关于在骨架检测之前操纵点云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！