人脸跟踪三维点云获取及三角面片分割显示（kinect2.0 SDK开发 ）

#include "stdafx.h"
#include "opencv2/opencv.hpp"
#include "Kinect.h"
#include "kinect.face.h"
#include "iostream"
#include "fstream"
using namespace cv;
using namespace std;

//安全释放指针
template<class Interface>
inline void SafeRelease(Interface *& pInterfaceToRelease)
{
	if (pInterfaceToRelease != NULL)
	{
		pInterfaceToRelease->Release();
		pInterfaceToRelease = NULL;
	}
}

//三角片数据结构
 struct TRIANGLE_DESC
{
	Point pt1, pt2, pt3;	
 };

//计算三角面片
vector<TRIANGLE_DESC> delaunayAlgorithm(const Rect boundRc, const vector<Point>& points)
{
	if (points.empty()||boundRc.width<=0||boundRc.height<=0)
	{
		return vector<TRIANGLE_DESC>();
	}
	vector<TRIANGLE_DESC> result;

	vector<Vec6f> _temp_result;
	Subdiv2D subdiv2d(boundRc);
	for (const auto point : points)
	{
		subdiv2d.insert(Point2f((float)point.x, (float)point.y));
	}
	subdiv2d.getTriangleList(_temp_result);

	for (const auto _tmp_vec : _temp_result)
	{
		Point pt1((int)_tmp_vec[0], (int)_tmp_vec[1]);
		Point pt2((int)_tmp_vec[2], (int)_tmp_vec[3]);
		Point pt3((int)_tmp_vec[4], (int)_tmp_vec[5]);
		if (boundRc.contains(pt1) && boundRc.contains(pt2) && boundRc.contains(pt3))  //去除最外头的三角形
		{
			TRIANGLE_DESC vec1;
			vec1.pt1 = pt1;
			vec1.pt2= pt2;
			vec1.pt3= pt3;
			result.push_back(vec1);
		}
		
	}
	return result;
}
//画三角面片
void drawTriangel(Mat &srcImg,Rect rectImg,vector<Point> pointsVec,int index)
{
	//每个模型的三角面片

	if (pointsVec.size() <= 2)  //防止画面片时报错
	{
		return ;
	}
	Rect rect1 = boundingRect(pointsVec);
	rect1 = rect1&rectImg;
	string label = format("Person %d", index);
	putText(srcImg, label, rect1.tl(), FONT_HERSHEY_SIMPLEX, 3.0f, Scalar(220, 0, 220), 4);
	vector<TRIANGLE_DESC> triangles = delaunayAlgorithm(rect1, pointsVec);
	rectangle(srcImg, rect1, Scalar(0, 0, 255), 5);
	for (int triangleCount = 0; triangleCount < triangles.size(); triangleCount++)
	{
		TRIANGLE_DESC triangle = triangles[triangleCount];
		line(srcImg, triangle.pt1, triangle.pt2, Scalar(255, 0, 0), 2);
		line(srcImg, triangle.pt2, triangle.pt3, Scalar(255, 0, 0), 2);
		line(srcImg, triangle.pt3, triangle.pt1, Scalar(255, 0, 0), 2);
	}//每个模型的三角面片
}

int _tmain(int argc, _TCHAR* argv[])
{

	IKinectSensor *kinectSensor;
	HRESULT hr = GetDefaultKinectSensor(&kinectSensor);
	kinectSensor->Open();

	//开启深度数据获取
	IDepthFrameSource* depths = nullptr;
	kinectSensor->get_DepthFrameSource(&depths);     // 获取深度数据源  
	int height, width;                               //图像宽和高
	IFrameDescription   * myDescription = nullptr;  //取得深度数据的分辨率
	depths->get_FrameDescription(&myDescription);
	myDescription->get_Height(&height);
	myDescription->get_Width(&width);
	myDescription->Release();

	IDepthFrameReader* depthRead = nullptr;
	depths->OpenReader(&depthRead); // 打开深度解析器  

	cout << "width " << width << "  height " << height << endl;
	Mat img16u(height, width, CV_16UC1);
	Mat img8u(height, width, CV_8UC1);

	//开启彩色数据获取
	IColorFrameSource* colorFramSouce = nullptr;
	kinectSensor->get_ColorFrameSource(&colorFramSouce); // 获取图像数据源  
	IColorFrameReader* colorFramRead = nullptr;
	colorFramSouce->OpenReader(&colorFramRead); // 打开解析器

	IFrameDescription* frameDs = nullptr;
	colorFramSouce->get_FrameDescription(&frameDs); // 获取信息的属性  
	int heightColor, widthColor;
	frameDs->get_Height(&heightColor);
	frameDs->get_Width(&widthColor);
	cout << "heightColor " << heightColor << "  widthColor " << widthColor << endl;
	Mat imgColor(heightColor, widthColor, CV_8UC4);

	//开启骨骼数据获取
	IBodyFrameSource    * myBodySource = nullptr;       //source
	kinectSensor->get_BodyFrameSource(&myBodySource);
	int heightBody, widthBody;


	IBodyFrameReader    * myBodyReader = nullptr;       //reader
	myBodySource->OpenReader(&myBodyReader);
	int myBodyCount = 0;
	myBodySource->get_BodyCount(&myBodyCount);

	ICoordinateMapper   * myMapper = nullptr;
	kinectSensor->get_CoordinateMapper(&myMapper);

	//高清脸部跟踪
	IHighDefinitionFaceFrameSource* hdfacesource[BODY_COUNT];
	IHighDefinitionFaceFrameReader* hdfacereader[BODY_COUNT];
	IFaceModelBuilder* facemodelbuilder[BODY_COUNT];  // 创建面部特征对齐
	bool produce[BODY_COUNT] = { false };
	IFaceAlignment* facealignment[BODY_COUNT];
	IFaceModel* facemodel[BODY_COUNT];
	vector<vector<float>> deformations(BODY_COUNT, vector<float>(FaceShapeDeformations_Count));

	for (int count = 0; count < BODY_COUNT; count++)
	{
		CreateHighDefinitionFaceFrameSource(kinectSensor, &hdfacesource[count]);
		hdfacesource[count]->OpenReader(&hdfacereader[count]);
		hdfacesource[count]->OpenModelBuilder(FaceModelBuilderAttributes_None, &facemodelbuilder[count]);
		facemodelbuilder[count]->BeginFaceDataCollection();
		CreateFaceAlignment(&facealignment[count]);
		CreateFaceModel(1.0f, FaceShapeDeformations_Count, &deformations[count][0], &facemodel[count]);
	}

	UINT32 vertex = 0;
	GetFaceModelVertexCount(&vertex);  //人脸顶点数量，1347

	while (true)
	{
		//获取深度摄像头数据
		IDepthFrame* depthFram = nullptr;
		if (depthRead->AcquireLatestFrame(&depthFram) == S_OK) //通过Reader尝试获取最新的一帧深度数据，放入深度帧中,并判断是否成功获取
		{
			depthFram->CopyFrameDataToArray(height * width, (UINT16 *)img16u.data); //先把数据存入16位的图像矩阵中
			img16u.convertTo(img8u, CV_8UC1, 255.0 / 4500);   //再把16位转换为8位
			imshow("深度图像", img8u);
			depthFram->Release();
		}
		//获取RGB摄像头数据
		IColorFrame* colorFram = nullptr;
		if (colorFramRead->AcquireLatestFrame(&colorFram) == S_OK) //通过Reader尝试获取最新的一帧深度数据，放入深度帧中,并判断是否成功获取
		{
			colorFram->CopyConvertedFrameDataToArray(heightColor*widthColor * 4, (BYTE*)imgColor.data, ColorImageFormat_Bgra);

			colorFram->Release();
		}
		//获取人骨胳数据
		IBodyFrame *myBodyFrame = nullptr;
		if (myBodyReader->AcquireLatestFrame(&myBodyFrame) == S_OK)
		{
			myBodyCount = 0;
			IBody   ** bodyArr = nullptr;
			myBodySource->get_BodyCount(&myBodyCount);
			bodyArr = new IBody *[myBodyCount];
			for (int i = 0; i < myBodyCount; i++)   //bodyArr的初始化
				bodyArr[i] = nullptr;

			if (myBodyFrame->GetAndRefreshBodyData(myBodyCount, bodyArr) == S_OK)
			{

				for (int i = 0; i < myBodyCount; i++)   //遍历6个人(可能用不完)
				{
					BOOLEAN     result = false;
					if (bodyArr[i]->get_IsTracked(&result) == S_OK && result)   //判断此人是否被侦测到
					{
						cout << "Body " << i << " tracked!" << endl;
						UINT64 trackingId = _UI64_MAX;
						if (bodyArr[i]->get_TrackingId(&trackingId) == S_OK)
						{
							hdfacesource[i]->put_TrackingId(trackingId);  //检测到的人体索映值传入
						}
					}
				}
			}

			for (int i = 0; i < myBodyCount; i++)
			{

				IHighDefinitionFaceFrame* hdfaceframe = nullptr;
				if (hdfacereader[i]->AcquireLatestFrame(&hdfaceframe) == S_OK && hdfaceframe != nullptr)
				{

					BOOLEAN tracked = false;
					if (SUCCEEDED(hdfaceframe->get_IsTrackingIdValid(&tracked)) && tracked)
					{

						if (SUCCEEDED(hdfaceframe->GetAndRefreshFaceAlignmentResult(facealignment[i])) && facealignment[i] != nullptr)
						{

							if (!produce[i])
							{
								FaceModelBuilderCollectionStatus collection;
								facemodelbuilder[i]->get_CollectionStatus(&collection);
								if (collection == FaceModelBuilderCollectionStatus::FaceModelBuilderCollectionStatus_Complete)
								{
									cout << "Status : Complete" << endl;
									putText(imgColor, "Status : Complete", Point(50, 50), FONT_HERSHEY_SIMPLEX, 1.0f, Scalar(0, 0, 255, 255), 2);
									IFaceModelData* facemodeldata = nullptr;
									if (SUCCEEDED(facemodelbuilder[i]->GetFaceData(&facemodeldata)) && facemodeldata != nullptr)
									{
										if (SUCCEEDED(facemodeldata->ProduceFaceModel(&facemodel[i])) && facemodel[i] != nullptr)
										{
											produce[i] = true;
										}
									}
									SafeRelease(facemodeldata);
								}

								vector<CameraSpacePoint> facePoints(vertex);  //facePoints(vertex)存的三维点 x,y,z坐标
								//vector<Point> points3Dto2D;//全部点有的不在平面上
								vector<Point> inImgrectPoints;//在图像内的点
								Rect rectImg = Rect(0, 0, widthColor, heightColor);

								if (SUCCEEDED(facemodel[i]->CalculateVerticesForAlignment(facealignment[i], vertex, &facePoints[0])))
								{
									ofstream outfile3D("Face3Ddata.txt");
									ofstream outfile2D("Face2Ddata.txt");
									//#pragma omp parallel for
									for (int point = 0; point < vertex; point+=1)
									{
										
										ColorSpacePoint colorSpacePoint;
										myMapper->MapCameraPointToColorSpace(facePoints[point], &colorSpacePoint);
										circle(imgColor, Point(colorSpacePoint.X, colorSpacePoint.Y), 4, Scalar(0, 255, 0, 255), -1);
										//points3Dto2D.push_back(Point(colorSpacePoint.X, colorSpacePoint.Y));  //注意：这里映射的点不一定在彩色图像上
										if (rectImg.contains(Point(colorSpacePoint.X, colorSpacePoint.Y)))
										{
											inImgrectPoints.push_back(Point(colorSpacePoint.X, colorSpacePoint.Y));
											outfile2D << colorSpacePoint.X << " \t" << colorSpacePoint.Y << endl;
										}
										outfile3D << facePoints[point].X << " \t" << facePoints[point].Y <<" \t"<< facePoints[point].Z << endl;
									}
									outfile2D.close();
									outfile3D.close();
									//画每个模型的三角面片
									//drawTriangel(imgColor,rectImg,inImgrectPoints,i);
									
								}
							}
						}
					}
				 }
			  }
			Mat imgShow;
			//imshow("原图", imgColor);
			resize(imgColor,imgShow,Size(800,480));
			imshow("彩色图", imgShow);
			myBodyFrame->Release();
	      }
		waitKey(1);
	  }
	depthRead->Release();        //释放不用的变量并且关闭感应器
	colorFramRead->Release();
	kinectSensor->Close();
	return 0;
}

这里注意下Kinect的开发环境配置，win8以上（我的win10），usb3.0接口

安装kinect2.0SDk，安装后一定重启，关机后再开似乎效果不一样

kinect.face.dll文件要复制到工程目录下

需要的库：