使用Kinect SDK进行头部跟踪？

Question

是否可以检测我是否用Kinect转动或倾斜我的头部 最好使用Kinect SDK。 我知道Forza 4会有一些头条新闻，但可以用SDK完成吗？

Answer 1

查看关于此类主题的第9频道tutorials。你会去参加Skeletal Fundamentals Video。但如果你想节省时间，这里有一些代码 的 XAML

<Window x:Class="SkeletalTracking.MainWindow"
    xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
    xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
    Title="MainWindow" Height="600" Width="800" Loaded="Window_Loaded" 
    xmlns:my="clr-namespace:Microsoft.Samples.Kinect.WpfViewers;assembly=Microsoft.Samples.Kinect.WpfViewers" 
    Closing="Window_Closing" WindowState="Maximized">       
<Canvas Name="MainCanvas">
    <my:KinectColorViewer Canvas.Left="0" Canvas.Top="0" Width="640" Height="480" Name="kinectColorViewer1" 
                          Kinect="{Binding ElementName=kinectSensorChooser1, Path=Kinect}" />
    <Ellipse Canvas.Left="0" Canvas.Top="0" Height="50" Name="leftEllipse"  Width="50" Fill="#FF4D298D" Opacity="1" Stroke="White" />
    <Ellipse Canvas.Left="100" Canvas.Top="0" Fill="#FF2CACE3" Height="50" Name="rightEllipse" Width="50" Opacity="1" Stroke="White" />
    <my:KinectSensorChooser Canvas.Left="250" Canvas.Top="380" Name="kinectSensorChooser1" Width="328" />
    <Image Canvas.Left="66" Canvas.Top="90" Height="87" Name="headImage" Stretch="Fill" Width="84" Source="/SkeletalTracking;component/c4f-color.png" />
</Canvas>

内部代码

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Windows;
using System.Windows.Controls;
using System.Windows.Data;
using System.Windows.Documents;
using System.Windows.Input;
using System.Windows.Media;
using System.Windows.Media.Imaging;
using System.Windows.Navigation;
using System.Windows.Shapes;
using Microsoft.Kinect;
using Coding4Fun.Kinect.Wpf; 

namespace SkeletalTracking
{
/// <summary>
/// Interaction logic for MainWindow.xaml
/// </summary>
public partial class MainWindow : Window
{
    public MainWindow()
    {
        InitializeComponent();
    }

    bool closing = false;
    const int skeletonCount = 6; 
    Skeleton[] allSkeletons = new Skeleton[skeletonCount];

    private void Window_Loaded(object sender, RoutedEventArgs e)
    {
        kinectSensorChooser1.KinectSensorChanged += new DependencyPropertyChangedEventHandler(kinectSensorChooser1_KinectSensorChanged);

    }

    void kinectSensorChooser1_KinectSensorChanged(object sender, DependencyPropertyChangedEventArgs e)
    {
        KinectSensor old = (KinectSensor)e.OldValue;

        StopKinect(old);

        KinectSensor sensor = (KinectSensor)e.NewValue;

        if (sensor == null)
        {
            return;
        }




        var parameters = new TransformSmoothParameters
        {
            Smoothing = 0.3f,
            Correction = 0.0f,
            Prediction = 0.0f,
            JitterRadius = 1.0f,
            MaxDeviationRadius = 0.5f
        };
        //sensor.SkeletonStream.Enable(parameters);

        sensor.SkeletonStream.Enable();

        sensor.AllFramesReady += new EventHandler<AllFramesReadyEventArgs>(sensor_AllFramesReady);
        sensor.DepthStream.Enable(DepthImageFormat.Resolution640x480Fps30); 
        sensor.ColorStream.Enable(ColorImageFormat.RgbResolution640x480Fps30);

        try
        {
            sensor.Start();
        }
        catch (System.IO.IOException)
        {
            kinectSensorChooser1.AppConflictOccurred();
        }
    }

    void sensor_AllFramesReady(object sender, AllFramesReadyEventArgs e)
    {
        if (closing)
        {
            return;
        }

        //Get a skeleton
        Skeleton first =  GetFirstSkeleton(e);

        if (first == null)
        {
            return; 
        }



        //set scaled position
        //ScalePosition(headImage, first.Joints[JointType.Head]);
        ScalePosition(leftEllipse, first.Joints[JointType.HandLeft]);
        ScalePosition(rightEllipse, first.Joints[JointType.HandRight]);

        GetCameraPoint(first, e); 

    }

    void GetCameraPoint(Skeleton first, AllFramesReadyEventArgs e)
    {

        using (DepthImageFrame depth = e.OpenDepthImageFrame())
        {
            if (depth == null ||
                kinectSensorChooser1.Kinect == null)
            {
                return;
            }


            //Map a joint location to a point on the depth map
            //head
            DepthImagePoint headDepthPoint =
                depth.MapFromSkeletonPoint(first.Joints[JointType.Head].Position);
            //left hand
            DepthImagePoint leftDepthPoint =
                depth.MapFromSkeletonPoint(first.Joints[JointType.HandLeft].Position);
            //right hand
            DepthImagePoint rightDepthPoint =
                depth.MapFromSkeletonPoint(first.Joints[JointType.HandRight].Position);


            //Map a depth point to a point on the color image
            //head
            ColorImagePoint headColorPoint =
                depth.MapToColorImagePoint(headDepthPoint.X, headDepthPoint.Y,
                ColorImageFormat.RgbResolution640x480Fps30);
            //left hand
            ColorImagePoint leftColorPoint =
                depth.MapToColorImagePoint(leftDepthPoint.X, leftDepthPoint.Y,
                ColorImageFormat.RgbResolution640x480Fps30);
            //right hand
            ColorImagePoint rightColorPoint =
                depth.MapToColorImagePoint(rightDepthPoint.X, rightDepthPoint.Y,
                ColorImageFormat.RgbResolution640x480Fps30);


            //Set location
            CameraPosition(headImage, headColorPoint);
            CameraPosition(leftEllipse, leftColorPoint);
            CameraPosition(rightEllipse, rightColorPoint);
        }        
    }


    Skeleton GetFirstSkeleton(AllFramesReadyEventArgs e)
    {
        using (SkeletonFrame skeletonFrameData = e.OpenSkeletonFrame())
        {
            if (skeletonFrameData == null)
            {
                return null; 
            }


            skeletonFrameData.CopySkeletonDataTo(allSkeletons);

            //get the first tracked skeleton
            Skeleton first = (from s in allSkeletons
                                     where s.TrackingState == SkeletonTrackingState.Tracked
                                     select s).FirstOrDefault();

            return first;

        }
    }

    private void StopKinect(KinectSensor sensor)
    {
        if (sensor != null)
        {
            if (sensor.IsRunning)
            {
                //stop sensor 
                sensor.Stop();

                //stop audio if not null
                if (sensor.AudioSource != null)
                {
                    sensor.AudioSource.Stop();
                }


            }
        }
    }

    private void CameraPosition(FrameworkElement element, ColorImagePoint point)
    {
        //Divide by 2 for width and height so point is right in the middle 
        // instead of in top/left corner
        Canvas.SetLeft(element, point.X - element.Width / 2);
        Canvas.SetTop(element, point.Y - element.Height / 2);

    }

    private void ScalePosition(FrameworkElement element, Joint joint)
    {
        //convert the value to X/Y
        //Joint scaledJoint = joint.ScaleTo(1280, 720); 

        //convert & scale (.3 = means 1/3 of joint distance)
        //Joint scaledJoint = joint.ScaleTo(1280, 720, .3f, .3f);

        Canvas.SetLeft(element, scaledJoint.Position.X);
        Canvas.SetTop(element, scaledJoint.Position.Y); 

    }


    private void Window_Closing(object sender, System.ComponentModel.CancelEventArgs e)
    {
        closing = true; 
        StopKinect(kinectSensorChooser1.Kinect); 
    }



   }
}

现在显然你的情况是你不希望在你的情况下进行手部跟踪，但我决定将其添加到那里。我个人会建议观看视频，因为他们解释了一切。祝你的项目好运！

Answer 2

Microsoft最近发布了Face Tracking API作为其最新SDK和其他工具包的一部分。下载此处链接的两个文件：

http://www.microsoft.com/en-us/kinectforwindows/develop/developer-downloads.aspx

包含了一些示例，但我发现它们有点难以集成到我自己的代码中。我发现这篇文章对于理解如何使用Face Tracking API更有用：

http://www.codeproject.com/Articles/394975/How-To-Use-Kinect-Face-Tracking-SDK

对于我的用例，MS Face Tracking无法正常工作。不要误会我的意思，这是优秀的面部跟踪，包括面部表情等，但它需要人比我们想要的更接近相机，并且它取决于照明条件，在我们的情况下，光线很差，因此跟踪充满了小跳跃和错误。我也发现了这种方法：

http://youtube.com/watch?v=AJHYxdTj3iQ&feature=related

虽然没有深入描述如何使用此方法，但它们确实提供了如何执行此操作的高级概述。它不跟踪面部表情......只跟踪头部方向。因此它不像MS Face Tracking那样功能齐全，但它似乎可以很好地进行头部方向跟踪，并且比大多数头部姿势估计技术更容易实现。