基于C++实现kinect+opencv 获取深度及彩色数据

开发环境 vs2010+OPENCV2.4.10

首先，下载最新的Kinect 2 SDK  http://www.microsoft.com/en-us/kinectforwindows/develop/downloads-docs.aspx

下载之后不要插入Kinect，最好也不用插入除了键盘鼠标以外的其它USB设备，然后安装SDK，安装完成之后插入Kinect，会有安装新设备的提示。安装完成之后可以去“开始”那里找到两个新安装的软件，一个是可以显示Kinect深度图，另外一个软件展示SDK中的各种例子程序。

进入SDK的安装目录，可以找到sample这个文件夹，里面是四种语言编写的例子，其中native是C++的，managed是C#的，还有另外两种语言不熟悉，我就熟悉C++，反正只是试试的，就用C++了。

opencv+kinect .cpp

#include <opencv2\opencv.hpp>
#include<iostream>
//windows的头文件，必须要，不然NuiApi.h用不了
#include <Windows.h>
//Kinect for windows 的头文件
#include "NuiApi.h"

using namespace std;
using namespace cv;

#include <d3d11.h>

//最远距离(mm)
const int MAX_DISTANCE = 3500;
//最近距离(mm)
const int MIN_DISTANCE = 200;

const LONG m_depthWidth = 640;
const LONG m_depthHeight = 480;
const LONG m_colorWidth = 640;
const LONG m_colorHeight = 480;
const LONG cBytesPerPixel = 4;

int main()
{
  //彩色图像
  Mat image_rgb;
  //深度图像
  Mat image_depth;

  //创建一个MAT
  image_rgb.create(480,640,CV_8UC3);
  image_depth.create(480,640,CV_8UC1);

  //一个KINECT实例指针
  INuiSensor* m_pNuiSensor = NULL;

  if (m_pNuiSensor != NULL)
  {
    return 0;
  }

  //记录当前连接KINECT的数量（为多连接做准备）
  int iSensorCount;
  //获得当前KINECT的数量
  HRESULT hr = NuiGetSensorCount(&iSensorCount);

  //按照序列初始化KINETC实例，这里就连接了一个KINECT，所以没有用到循环
  hr = NuiCreateSensorByIndex(iSensorCount - 1, &m_pNuiSensor);
  //初始化，让其可以接收彩色和深度数据流
  hr = m_pNuiSensor->NuiInitialize(NUI_INITIALIZE_FLAG_USES_COLOR | NUI_INITIALIZE_FLAG_USES_DEPTH);

  //判断是否出错
  if (FAILED(hr))
  {
    cout<<"NuiInitialize failed"<<endl;
    return hr;
  }

  //彩色图像获取下一帧事件
  HANDLE nextColorFrameEvent = CreateEvent(NULL, TRUE, FALSE, NULL);
  //彩色图像事件句柄
  HANDLE colorStreamHandle = NULL;
  //深度图像获取下一帧事件
  HANDLE nextDepthFrameEvent = CreateEvent(NULL, TRUE, FALSE, NULL);
  //深度图像事件句柄
  HANDLE depthStreamHandle = NULL;

  //实例打开数据流，这里NUI_IMAGE_TYPE_COLOR表示彩色图像
  hr = m_pNuiSensor->NuiImageStreamOpen(NUI_IMAGE_TYPE_COLOR, NUI_IMAGE_RESOLUTION_640x480, 0,2,nextColorFrameEvent,&colorStreamHandle);

  if( FAILED( hr ) )//判断是否提取正确
  {
    cout<<"Could not open color image stream video"<<endl;
    m_pNuiSensor->NuiShutdown();
    return hr;
  }

  //实例打开数据流，这里NUI_IMAGE_TYPE_DEPTH表示深度图像
  hr = m_pNuiSensor->NuiImageStreamOpen(NUI_IMAGE_TYPE_DEPTH, NUI_IMAGE_RESOLUTION_640x480, 0,2, nextDepthFrameEvent, &depthStreamHandle);

  if( FAILED( hr ) )//判断是否提取正确
  {
    cout<<"Could not open color image stream video"<<endl;
    m_pNuiSensor->NuiShutdown();
    return hr;
  }

  cv::namedWindow("depth", CV_WINDOW_AUTOSIZE);
  moveWindow("depth",300,600);
  cv::namedWindow("colorImage",CV_WINDOW_AUTOSIZE);
  moveWindow("colorImage",0,200);

  while (1)
  {
    NUI_IMAGE_FRAME pImageFrame_rgb;
    NUI_IMAGE_FRAME pImageFrame_depth;

    //无限等待新的彩色数据，等到后返回
    if (WaitForSingleObject(nextColorFrameEvent, 0) == 0)
    {
      //从刚才打开数据流的流句柄中得到该帧数据，读取到的数据地址存于pImageFrame
      hr = m_pNuiSensor->NuiImageStreamGetNextFrame(colorStreamHandle, 0, &pImageFrame_rgb);
      if (FAILED(hr))
      {
        cout<<"Could not get color image"<<endl;
        m_pNuiSensor->NuiShutdown();
        return -1;
      }

      INuiFrameTexture *pTexture = pImageFrame_rgb.pFrameTexture;
      NUI_LOCKED_RECT lockedRect;

      //提取数据帧到LockedRect，它包括两个数据对象：pitch每行字节数，pBits第一个字节地址
      //并锁定数据，这样当我们读数据的时候，kinect就不会去修改它

      pTexture->LockRect(0, &lockedRect, NULL, 0);
      //确认获得的数据是否有效
      if (lockedRect.Pitch != 0)
      {
        //将数据转换为OpenCV的Mat格式
        for (int i = 0; i < image_rgb.rows; i++)
        {
          //第i行的指针
          uchar *prt = image_rgb.ptr(i);

          //每个字节代表一个颜色信息，直接使用uchar
          uchar *pBuffer = (uchar*)(lockedRect.pBits) + i * lockedRect.Pitch;

          for (int j = 0; j < image_rgb.cols; j++)
          {
            prt[3 * j] = pBuffer[4 * j];//内部数据是4个字节，0-1-2是BGR，第4个现在未使用
            prt[3 * j + 1] = pBuffer[4 * j + 1];
            prt[3 * j + 2] = pBuffer[4 * j + 2];
          }
        }

        imshow("colorImage",image_rgb);
        //解除锁定
        pTexture->UnlockRect(0);
        //释放帧
        m_pNuiSensor->NuiImageStreamReleaseFrame(colorStreamHandle, &pImageFrame_rgb );
      }
      else
      {
        cout<<"Buffer length of received texture is bogus\r\n"<<endl;
      }

      BOOL nearMode;
      INuiFrameTexture* pColorToDepthTexture; 

      //深度图像的处理
      if (WaitForSingleObject(nextDepthFrameEvent, INFINITE) == 0)
      {

        hr = m_pNuiSensor->NuiImageStreamGetNextFrame(depthStreamHandle, 0 , &pImageFrame_depth);

        if (FAILED(hr))
        {
          cout<<"Could not get color image"<<endl;
          NuiShutdown();
          return -1;
        }

        hr = m_pNuiSensor->NuiImageFrameGetDepthImagePixelFrameTexture(
          depthStreamHandle, &pImageFrame_depth, &nearMode, &pColorToDepthTexture);
        INuiFrameTexture *pTexture = pImageFrame_depth.pFrameTexture;
        NUI_LOCKED_RECT lockedRect;
        NUI_LOCKED_RECT ColorToDepthLockRect; 

        pTexture->LockRect(0, &lockedRect, NULL, 0);
        pColorToDepthTexture->LockRect(0,&ColorToDepthLockRect,NULL,0); 

        //归一化
        for (int i = 0; i < image_depth.rows; i++)
        {
          uchar *prt = image_depth.ptr<uchar>(i);

          uchar* pBuffer = (uchar*)(lockedRect.pBits) + i * lockedRect.Pitch;
          //这里需要转换，因为每个深度数据是2个字节，应将BYTE转成USHORT
          USHORT *pBufferRun = (USHORT*)pBuffer;

          for (int j = 0; j < image_depth.cols; j++)
          {
            //先向，将数据归一化处理，对深度距离在300mm-3500mm范围内的像素，映射到【0—255】内，
            //超出范围的，都去做是边缘像素
            if (pBufferRun[j] << 3 > MAX_DISTANCE) prt[j] = 255;
            else if(pBufferRun[j] << 3 < MIN_DISTANCE) prt[j] = 0;
            else prt[j] = (BYTE)(256 * (pBufferRun[j] << 3)/ MAX_DISTANCE);
          }
        }
        imshow("depth", image_depth);

        //接下来是对齐部分，将前景抠出来

        //存放深度点的参数
        NUI_DEPTH_IMAGE_POINT* depthPoints = new NUI_DEPTH_IMAGE_POINT[640 * 480];
        if (ColorToDepthLockRect.Pitch != 0)
        {
          HRESULT hrState = S_OK; 

          //一个能在不同空间坐标转变的类（包括：深度，彩色，骨骼）
          INuiCoordinateMapper* pMapper; 

          //设置KINECT实例的空间坐标系
          hrState = m_pNuiSensor->NuiGetCoordinateMapper(&pMapper); 

          if (FAILED(hrState))
          {
            return hrState;
          } 

          //重要的一步：从颜色空间映射到深度空间。参数说明：
          //【参数1】：彩色图像的类型
          //【参数2】：彩色图像的分辨率
          //【参数3】：深度图像的分辨率
          //【参数4】：深度图像的个数
          //【参数5】：深度像素点数
          //【参数6】：取内存的大小，个数。类型为NUI_DEPTH_IMAGE_PIXEL
          //【参数7】：存放映射结果点的参数
          hrState = pMapper->MapColorFrameToDepthFrame(NUI_IMAGE_TYPE_COLOR, NUI_IMAGE_RESOLUTION_640x480, NUI_IMAGE_RESOLUTION_640x480,
            640 * 480, (NUI_DEPTH_IMAGE_PIXEL*)ColorToDepthLockRect.pBits,640 * 480, depthPoints); 

          if (FAILED(hrState))
          {
            return hrState;
          } 

          //显示的图像
          Mat show;
          show.create(480,640,CV_8UC3);
          show = 0;

          for (int i = 0; i < image_rgb.rows; i++)
          {
            for (int j = 0; j < image_rgb.cols; j++)
            {
              uchar *prt_rgb = image_rgb.ptr(i);
              uchar *prt_show = show.ptr(i);
              //在内存中偏移量
              long index = i * 640 + j;
              //从保存了映射坐标的数组中获取点
              NUI_DEPTH_IMAGE_POINT depthPointAtIndex = depthPoints[index]; 

              //边界判断
              if (depthPointAtIndex.x >= 0 && depthPointAtIndex.x < image_depth.cols &&
                depthPointAtIndex.y >=0 && depthPointAtIndex.y < image_depth.rows)
              {
                //深度判断，在MIN_DISTANCE与MAX_DISTANCE之间的当成前景，显示出来
                //这个使用也很重要，当使用真正的深度像素点再在深度图像中获取深度值来判断的时候，会出错
                if (depthPointAtIndex.depth >= MIN_DISTANCE && depthPointAtIndex.depth <= MAX_DISTANCE)
                {
                  prt_show[3 * j]   = prt_rgb[j * 3];
                  prt_show[3 * j + 1] = prt_rgb[j * 3 + 1];
                  prt_show[3 * j + 2] = prt_rgb[j * 3 + 2];
                }
              }
            }
          }
          imshow("show", show);
        }

        delete []depthPoints;

        pTexture->UnlockRect(0);
        m_pNuiSensor->NuiImageStreamReleaseFrame(depthStreamHandle, &pImageFrame_depth);
      }

      else
      {
        cout<<"Buffer length of received texture is bogus\r\n"<<endl;
      }
    }

    if (cvWaitKey(20) == 27)
      break;
  }
  return 0;
}