感兴趣区域的问题Kinect SDK 1.7 [英] Problems with the region of interest Kinect SDK 1.7

问题描述

您好，

我正在尝试在kinect SDK 1.7中跟踪一张脸而不跟踪Skeleton，我读了一篇帖子，其中autor更改了Microsoft的一些代码。 Kinect.Toolkit.FaceTracker示例为了允许我们将RegionOfInterest和
Skeleton作为null调用StartTracking方法，并将HeadPoints的continueTracking作为null。它适用于我，我跟踪没有骨架的脸，但它的工作太慢。视频跳过许多帧，我无法在Kinect工作的同时进行任何其他操作。我需要
来使用感兴趣的区域（Rect roi），其中kinect必须搜索要跟踪的面，但如果我用方法调用方法track（）它不起作用，我显示矩形要知道我在哪里"放"我的脸。这篇帖子说明了如何用彩色和深度图像跟踪
a面部是  this 。第二个回复。我希望有人可以帮助我。
感谢大家。

PS：请原谅我糟糕的英语。

解决方案

由于RegionOfInterest仅在StartTracking函数中使用过一次，因此主要问题是headPoints。

我追踪了一张没有骨架的脸，但效果太慢了。

这是不使用headPoints参数的主要副作用。没有这些值，您可以稍微减少总面积，但是您仍在处理比您更大的区域如果你知道头位置在哪里。

您使用的是C＃Toolkit api或C ++ api吗？您计算的RegionOfInterest是否包含足够的所有科目的头部和颈部区域？

IFTFaceTracker :: StartTracking / ContinueTracking

http://msdn.microsoft.com/en-us/library/microsoft.kinect.facetracking .iftfacetracker.starttracking.aspx
$
http://msdn.microsoft.com/en-us/library/microsoft.kinect.facetracking.iftfacetracker.continuetracking.aspx

Hello,

I'm trying to track a face in kinect SDK 1.7 without tracking a Skeleton, I read a post in which the autor changes some code in Microsoft.Kinect.Toolkit.FaceTracker sample in order to allow us to call the StartTracking method with the RegionOfInterest and Skeleton as null and continuesTracking with HeadPoints as null. It works for me and I tracked a face without a skeleton but it works too slow. The video skips many frames and I can't do any other operations at the same time as the Kinect is working. I need to use a region of interest (Rect roi) in where the kinect must search the face to track but if I call the method track() with a rect it doesn't works, I'm showing the rectangle to know where have I to "put" my face. The post which says how to track a face only with color and depth image is this. The second reply. I hope that someone can help me. Thanks to everyone.

PS: Excuse my bad english.

解决方案

Since RegionOfInterest is only used once in the StartTracking function, the main issue are the headPoints.

I tracked a face without a skeleton but it works too slow.

This is the main side effect of not using the headPoints parameter. Without these values, you can minimize the total area a bit, but you are still processing a larger area than you would if you know where the head positions are.

Are you using the C# Toolkit api's or C++ api's? Does the RegionOfInterest you calculated encompass enough area for the head and neck for all subjects?

IFTFaceTracker::StartTracking/ContinueTracking
http://msdn.microsoft.com/en-us/library/microsoft.kinect.facetracking.iftfacetracker.starttracking.aspx
http://msdn.microsoft.com/en-us/library/microsoft.kinect.facetracking.iftfacetracker.continuetracking.aspx

这篇关于感兴趣区域的问题Kinect SDK 1.7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！