Kinect for Windows SDK 1.5 引入 Microsoft.Kinect.Toolkit.FaceTracking 命名空間，提供原生托管級人臉追蹤能力，無需第三方OpenCV/EmguCV即可實現 121 點三維面部網格、動畫單元係數、三角形網格提取。本文通過構建一個WPF「虛擬臉部」Demo，完整演示：環境配置、三流同步獲取、FaceTracker 核心調用、關鍵數據結構解析、XAML矢量繪製臉部、動畫單元駆動表情變化的全流程。
Microsoft.Kinect.dllMicrosoft.Kinect.Toolkit.dllMicrosoft.Kinect.Toolkit.FaceTracking.dll（Toolkit Samples\bin 目錄）bin/Debug）：
FaceTrackData.dllFaceTrackLib.dll（x86/x64 對應版本）
// 啟用三大流：彩色、深度（低分辨率高幂率）、骨骼
kinectSensor.ColorStream.Enable();
kinectSensor.DepthStream.Enable(DepthImageFormat.Resolution80x60Fps30); // 低分辨率提升FPS
kinectSensor.SkeletonStream.TrackingMode = SkeletonTrackingMode.Seated; // 坐姿模式適配近景
kinectSensor.SkeletonStream.Enable(new TransformSmoothParameters {
Correction = 0.5f, JitterRadius = 0.05f, MaxDeviationRadius = 0.05f, Prediction = 0.5f, Smoothing = 0.5f
});
// 緩衝區預分配
colorPixelData = new byte[kinectSensor.ColorStream.FramePixelDataLength];
depthPixelData = new short[kinectSensor.DepthStream.FramePixelDataLength];
skeletonData = new Skeleton[6];
// 事件訂閱與啟動
kinectSensor.AllFramesReady += kinectSensor_AllFramesReady;
kinectSensor.Start();
// 初始化 FaceTracker（核心對象）
faceTracker = new FaceTracker(kinectSensor);
void kinectSensor_AllFramesReady(object sender, AllFramesReadyEventArgs e) {
// 1. 獲取三幂數據
using (ColorImageFrame colorFrame = e.OpenColorImageFrame())
using (DepthImageFrame depthFrame = e.OpenDepthImageFrame())
using (SkeletonFrame skeletonFrame = e.OpenSkeletonFrame()) {
if (colorFrame == null || depthFrame == null || skeletonFrame == null) return;
colorFrame.CopyPixelDataTo(colorPixelData);
depthFrame.CopyPixelDataTo(depthPixelData);
skeletonFrame.CopySkeletonDataTo(skeletonData);
}
// 2. 取首個被追蹤骨骼
var skeleton = skeletonData.FirstOrDefault(s => s.TrackingState == SkeletonTrackingState.Tracked);
if (skeleton == null) return;
// 3. FaceTracker 處理，返回 FaceTrackFrame
FaceTrackFrame faceFrame = faceTracker.Track(
kinectSensor.ColorStream.Format, colorPixelData,
kinectSensor.DepthStream.Format, depthPixelData,
skeleton
);
// 4. 後續處理見下文...
}
float[]，索引對應 AnimationUnit 枚舉（JawLower下顎張開、LipStretcher嘴唇拉伸、BrowLower眉頭降低、BrowRaiser眉毛抬起等），取值範圍 [-1, 1]Get3DShape() 構建三維面部網格在 Canvas 中組合 Ellipse（臉輪廓、眼睛、嘴巴）與 Rectangle（眉毛），利用 RenderTransform（ScaleTransform、TranslateTransform、RotateTransform）實現變形：
<Canvas x:Name="MainCanvas" Width="500" Height="500">
<Canvas.RenderTransform>
<TransformGroup>
<RotateTransform x:Name="CanvasRotate" CenterX="250" CenterY="250" />
</TransformGroup>
</Canvas.RenderTransform>
<Ellipse x:Name="Face" Width="300" Height="300" Stroke="Black" StrokeThickness="2" Canvas.Left="105" Canvas.Top="6" />
<Ellipse x:Name="LeftEye" Width="30" Height="30" Stroke="Black" StrokeThickness="2" Canvas.Left="289" Canvas.Top="102" />
<Ellipse x:Name="RightEye" Width="30" Height="30" Stroke="Black" StrokeThickness="2" Canvas.Left="194" Canvas.Top="102" />
<Ellipse x:Name="Mouth" Width="64" Height="18" Stroke="Black" StrokeThickness="2" Canvas.Left="224" Canvas.Top="239">
<Ellipse.RenderTransform>
<ScaleTransform x:Name="MouthScaleTransform" CenterX="32" CenterY="9" ScaleX="1" ScaleY="1" />
</Ellipse.RenderTransform>
</Ellipse>
<!-- 眉毛用 Rectangle 模擬，通過 TranslateTransform(Y軸位移) + RotateTransform(角度) 表達 -->
<Rectangle x:Name="RightBrow" Width="70" Height="5" Fill="Black" Canvas.Left="169" Canvas.Top="80">
<Rectangle.RenderTransform>
<TransformGroup>
<TranslateTransform x:Name="RightBrow" />
<RotateTransform x:Name="RightBrowRotate" CenterX="50" Angle="0" />
</TransformGroup>
</Rectangle.RenderTransform>
</Rectangle>
<Rectangle x:Name="LeftBrow" Width="70" Height="5" Fill="Black" Canvas.Left="274" Canvas.Top="80"> ... </Rectangle>
</Canvas>
if (faceFrame.TrackSuccessful) {
var AUCoeff = faceFrame.GetAnimationUnitCoefficients();
// 嘴巴張合：JawLower (下顎張開) 控制 ScaleY，LipStretcher (嘴唇拉伸) 控制 ScaleX
var jawLowerer = AUCoeff[(int)AnimationUnit.JawLower];
jawLowerer = jawLowerer < 0 ? 0 : jawLowerer; // 修正負值
MouthScaleTransform.ScaleY = jawLowerer * 5 + 0.1; // 放大係數並保留基礎高度
MouthScaleTransform.ScaleX = AUCoeff[(int)AnimationUnit.LipStretcher] + 1;
// 眉毛：BrowLower (眉頭降低) 控制 Y 軸位移，BrowRaiser (眉毛抬起) 控制旋轉角度
var browLower = AUCoeff[(int)AnimationUnit.BrowLower] * 40;
LeftBrow.Y = RightBrow.Y = browLower;
var browRaiser = AUCoeff[(int)AnimationUnit.BrowRaiser] * 20;
RightBrowRotate.Angle = browRaiser;
LeftBrowRotate.Angle = -browRaiser; // 左右眉毛旋轉方向相反
// 頭部旋轉：直接應用 FaceTrackFrame.Rotation.Z
CanvasRotate.Angle = faceFrame.Rotation.Z;
}
運行效果截圖（張嘴、皺眉、抬眉、轉頭）：
Get3DShape() + GetTriangles() 構建 MeshGeometry3D，配合 Viewport3D 渲染動態三維臉模（Toolkit範例 FaceTracking3D 演示）GetProjected3DShape() 關鍵點（瞳孔、嘴角、鼻尖）精準疊加AR素材JawLower（打哈欠）、EyeClosure（眨眼頻率/閉眼時長）、頭部姿態偏移SDK 內建 FaceTracking 以「三流融合輸入 → 統一 FaceTracker 處理 → 標準化數據結構輸出」的極簡鏈路，將工業級面部捕捉能力降維為托管API。本文Demo僅用 100 行核心代碼實現了從數據獲取到表情駆動的閉環，證明其極高的集成效率。開發者可直接在此基礎上擴展 3D重建、表情計算、AR交互等高階應用。
作者：yangecnu | 出處：https://www.cnblogs.com/yangecnu/
