camera calibration 摄像头标定 代码整理(2007-2-6 修正)

这里介绍了一个使用directshow标定与Matlab的示例,写的很详细。

 http://w3.impa.br/~pcezar/3dp/original/CVL_html/appPage/doc_calib.html#Tutorial

 

OPENCV没有提供完整的示例,自己整理了一下,贴出来记录。

步骤如下:

首先自制一张标定图片,用A4纸打印出来,设定距离,再设定标定棋盘的格子数目,如8×6,以下是我做的图片8×8


然后利用cvFindChessboardCorners找到棋盘在摄像头中的2D位置,这里cvFindChessboardCorners不太稳定,有时不能工作。

计算实际的距离,应该是3D的距离。我设定为21.6毫米,既在A4纸上为两厘米。

再用cvCalibrateCamera2计算内参,

最后用cvUndistort2纠正图像的变形。

结果如下:

 

源代码如下:

 

#include  " stdafx.h "
#include 
< stdio.h >
#include 
< stdlib.h >
#include 
< string .h >
//  OpenCV
#include  < cxcore.h >
#include 
< cv.h >
#include 
< highgui.h >
#include 
< cvaux.h >


void  InitCorners3D(CvMat  * Corners3D, CvSize ChessBoardSize,  int  Nimages,  float  SquareSize);
void  makeChessBoard();
int  myFindChessboardCorners(  const   void *  image, CvSize pattern_size,
                             CvPoint2D32f
*  corners,  int *  corner_count = NULL,
                             
int  flags = CV_CALIB_CB_ADAPTIVE_THRESH );


inline 
int  drawCorssMark(IplImage  * dst,CvPoint pt)
/*************************************************
  Function:        main_loop
  Description:     绘制一个十字标记                    
  Calls:          
  Called By:      
  Input:           RGB image,  pt               
  Output:         
  Return:         
  Others:          需要检查坐标是否越界 to do list
************************************************
*/

{
    
    
const int cross_len = 4;
    CvPoint pt1,pt2,pt3,pt4;
    pt1.x 
= pt.x;
    pt1.y 
= pt.y - cross_len;
    pt2.x 
= pt.x;
    pt2.y 
= pt.y + cross_len;
    pt3.x 
= pt.x - cross_len;
    pt3.y 
= pt.y;
    pt4.x 
= pt.x + cross_len;
    pt4.y 
= pt.y;

    cvLine(dst,pt1,pt2,CV_RGB(
0,255,0),2,CV_AA, 0 );    
    cvLine(dst,pt3,pt4,CV_RGB(
0,255,0),2,CV_AA, 0 );
    
    
return 0;
}


/* declarations for OpenCV */
IplImage                 
* current_frame_rgb,grid;
IplImage                 
* current_frame_gray;
IplImage                 
* chessBoard_Img;

int                        Thresholdness  =   120 ;

int  image_width  =   320 ;
int  image_height  =   240 ;

bool  verbose  =   false ;

const   int  ChessBoardSize_w  =   7 ;
const   int  ChessBoardSize_h  =   7 ;
//  Calibration stuff
bool             calibration_done  =   false ;
const  CvSize     ChessBoardSize  =  cvSize(ChessBoardSize_w,ChessBoardSize_h);
// float             SquareWidth = 21.6f;  // 实际距离 毫米单位 在A4纸上为两厘米
float              SquareWidth  =   17 // 投影实际距离 毫米单位  200

const     int  NPoints  =  ChessBoardSize_w * ChessBoardSize_h;
const     int  NImages  =   20 // Number of images to collect 

CvPoint2D32f corners[NPoints
* NImages];
int  corner_count[NImages]  =   {0} ;
int  captured_frames  =   0 ;

CvMat 
* intrinsics;
CvMat 
* distortion_coeff;
CvMat 
* rotation_vectors;
CvMat 
* translation_vectors;
CvMat 
* object_points;
CvMat 
* point_counts;
CvMat 
* image_points;
int  find_corners_result  = 0  ;


void  on_mouse(  int   event int  x,  int  y,  int  flags,  void *  param )
{

    
ifevent == CV_EVENT_LBUTTONDOWN )
    
{
        
//calibration_done = true; 
    }

}



int  main( int  argc,  char   * argv[])
{


  CvFont font;
  cvInitFont( 
&font, CV_FONT_VECTOR0,55078);
  
  intrinsics         
= cvCreateMat(3,3,CV_32FC1);
  distortion_coeff     
= cvCreateMat(1,4,CV_32FC1);
  rotation_vectors     
= cvCreateMat(NImages,3,CV_32FC1);
  translation_vectors     
= cvCreateMat(NImages,3,CV_32FC1);

  point_counts         
= cvCreateMat(NImages,1,CV_32SC1);

  object_points     
= cvCreateMat(NImages*NPoints,3,CV_32FC1);
  image_points         
= cvCreateMat(NImages*NPoints,2,CV_32FC1);


  
// Function to fill in the real-world points of the checkerboard
  InitCorners3D(object_points, ChessBoardSize, NImages, SquareWidth);


  CvCapture
* capture = 0;

  
  
if( argc == 1 || (argc == 2 && strlen(argv[1]) == 1 && isdigit(argv[1][0])))
      capture 
= cvCaptureFromCAM( argc == 2 ? argv[1][0- '0' : 0 );
  
else if( argc == 2 )
      capture 
= cvCaptureFromAVI( argv[1] );

  
if!capture )
  
{
      fprintf(stderr,
"Could not initialize capturing... ");
      
return -1;
  }

  
  
  
// Initialize all of the IplImage structures
  current_frame_rgb = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 3);
  
  IplImage 
*current_frame_rgb2 = cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 3);
  current_frame_gray 
= cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 1);

  chessBoard_Img   
= cvCreateImage(cvSize(image_width, image_height), IPL_DEPTH_8U, 3);  
  current_frame_rgb2
->origin = chessBoard_Img->origin  = current_frame_gray->origin = current_frame_rgb->origin = 1;
  
  makeChessBoard();

  cvNamedWindow( 
"result"0);
  cvNamedWindow( 
"Window 0"0);
  cvNamedWindow( 
"grid"0);
  cvMoveWindow( 
"grid"100,100);
  cvSetMouseCallback( 
"Window 0", on_mouse, 0 );  
  cvCreateTrackbar(
"Thresholdness","Window 0",&Thresholdness, 255,0);
  
  
while (!calibration_done)
  
{

    
while (captured_frames < NImages)
    
{
      current_frame_rgb 
= cvQueryFrame( capture );
      
//current_frame_rgb = cvLoadImage( "c:/BoardStereoL3.jpg" );
      
//cvCopy(chessBoard_Img,current_frame_rgb);

      
if!current_frame_rgb )
          
break;

      cvCopy(current_frame_rgb,current_frame_rgb2);
      cvCvtColor(current_frame_rgb, current_frame_gray, CV_BGR2GRAY);
      
//cvThreshold(current_frame_gray,current_frame_gray,Thresholdness,255,CV_THRESH_BINARY);
      
//cvThreshold(current_frame_gray,current_frame_gray,150,255,CV_THRESH_BINARY_INV);

/*
    int pos = 1;
    IplConvKernel* element = 0;
    const int element_shape = CV_SHAPE_ELLIPSE;
    element = cvCreateStructuringElementEx( pos*2+1, pos*2+1, pos, pos, element_shape, 0 );
    cvDilate(current_frame_gray,current_frame_gray,element,1);
    cvErode(current_frame_gray,current_frame_gray,element,1);
    cvReleaseStructuringElement(&element);
*/

    
    find_corners_result 
= cvFindChessboardCorners(current_frame_gray,
                                          ChessBoardSize,
                                          
&corners[captured_frames*NPoints],
                                          
&corner_count[captured_frames],
                                          
0);



    cvDrawChessboardCorners(current_frame_rgb2, ChessBoardSize, 
&corners[captured_frames*NPoints], NPoints, find_corners_result);


    cvShowImage(
"Window 0",current_frame_rgb2);
    cvShowImage(
"grid",chessBoard_Img);

    
if(find_corners_result==1)
    
{
        cvWaitKey(
2000);
        cvSaveImage(
"c:/hardyinCV.jpg",current_frame_rgb2);
        captured_frames
++;
    }

    
//cvShowImage("result",current_frame_gray);

    intrinsics
->data.fl[0= 256.8093262;   //fx        
    intrinsics->data.fl[2= 160.2826538;   //cx
    intrinsics->data.fl[4= 254.7511139;   //fy
    intrinsics->data.fl[5= 127.6264572;   //cy

    intrinsics
->data.fl[1= 0;   
    intrinsics
->data.fl[3= 0;   
    intrinsics
->data.fl[6= 0;   
    intrinsics
->data.fl[7= 0;   
    intrinsics
->data.fl[8= 1;       

    distortion_coeff
->data.fl[0= -0.193740;  //k1
    distortion_coeff->data.fl[1= -0.378588;  //k2
    distortion_coeff->data.fl[2= 0.028980;   //p1
    distortion_coeff->data.fl[3= 0.008136;   //p2

    cvWaitKey(
40);
    find_corners_result 
= 0;
    }
   
    
//if (find_corners_result !=0)
    {

        printf(
" ");

        cvSetData( image_points, corners, 
sizeof(CvPoint2D32f));
        cvSetData( point_counts, 
&corner_count, sizeof(int));

        
        cvCalibrateCamera2( object_points,
            image_points,
            point_counts,
            cvSize(image_width,image_height),
            intrinsics,
            distortion_coeff,
            rotation_vectors,
            translation_vectors,
            
0);
        

        
// [fx 0 cx; 0 fy cy; 0 0 1].
        cvUndistort2(current_frame_rgb,current_frame_rgb,intrinsics,distortion_coeff);
        cvShowImage(
"result",current_frame_rgb);


        
float intr[3][3= {0.0};
        
float dist[4= {0.0};
        
float tranv[3= {0.0};
        
float rotv[3= {0.0};

        
for ( int i = 0; i < 3; i++)
        
{
            
for ( int j = 0; j < 3; j++)
            
{
                intr[i][j] 
= ((float*)(intrinsics->data.ptr + intrinsics->step*i))[j];
            }

            dist[i] 
= ((float*)(distortion_coeff->data.ptr))[i];
            tranv[i] 
= ((float*)(translation_vectors->data.ptr))[i];
            rotv[i] 
= ((float*)(rotation_vectors->data.ptr))[i];
        }

        dist[
3= ((float*)(distortion_coeff->data.ptr))[3];

        printf(
"----------------------------------------- ");
        printf(
"INTRINSIC MATRIX:  ");
        printf(
"[ %6.4f %6.4f %6.4f ]  ", intr[0][0], intr[0][1], intr[0][2]);
        printf(
"[ %6.4f %6.4f %6.4f ]  ", intr[1][0], intr[1][1], intr[1][2]);
        printf(
"[ %6.4f %6.4f %6.4f ]  ", intr[2][0], intr[2][1], intr[2][2]);
        printf(
"----------------------------------------- ");
        printf(
"DISTORTION VECTOR:  ");
        printf(
"[ %6.4f %6.4f %6.4f %6.4f ]  ", dist[0], dist[1], dist[2], dist[3]);
        printf(
"----------------------------------------- ");
        printf(
"ROTATION VECTOR:  ");
        printf(
"[ %6.4f %6.4f %6.4f ]  ", rotv[0], rotv[1], rotv[2]);
        printf(
"TRANSLATION VECTOR:  ");
        printf(
"[ %6.4f %6.4f %6.4f ]  ", tranv[0], tranv[1], tranv[2]);
        printf(
"----------------------------------------- ");

        cvWaitKey(
0);

        calibration_done 
= true;      
    }

               
  }


  exit(
0);
  cvDestroyAllWindows();
}


void  InitCorners3D(CvMat  * Corners3D, CvSize ChessBoardSize,  int  NImages,  float  SquareSize)
{
  
int CurrentImage = 0;
  
int CurrentRow = 0;
  
int CurrentColumn = 0;
  
int NPoints = ChessBoardSize.height*ChessBoardSize.width;
  
float * temppoints = new float[NImages*NPoints*3];

  
// for now, assuming we're row-scanning
  for (CurrentImage = 0 ; CurrentImage < NImages ; CurrentImage++)
  
{
    
for (CurrentRow = 0; CurrentRow < ChessBoardSize.height; CurrentRow++)
    
{
      
for (CurrentColumn = 0; CurrentColumn < ChessBoardSize.width; CurrentColumn++)
      
{
          temppoints[(CurrentImage
*NPoints*3)+(CurrentRow*ChessBoardSize.width + CurrentColumn)*3]=(float)CurrentRow*SquareSize;
          temppoints[(CurrentImage
*NPoints*3)+(CurrentRow*ChessBoardSize.width + CurrentColumn)*3+1]=(float)CurrentColumn*SquareSize;
          temppoints[(CurrentImage
*NPoints*3)+(CurrentRow*ChessBoardSize.width + CurrentColumn)*3+2]=0.f;
      }

    }

  }

  (
*Corners3D) = cvMat(NImages*NPoints,3,CV_32FC1, temppoints);
}


int  myFindChessboardCorners(  const   void *  image, CvSize pattern_size,
                             CvPoint2D32f
*  corners,  int *  corner_count,
                             
int  flags )

{


    IplImage
* eig = cvCreateImage( cvGetSize(image), 321 );
    IplImage
* temp = cvCreateImage( cvGetSize(image), 321 );
    
double quality = 0.01;
    
double min_distance = 5;
    
int win_size =10;

    
int count = pattern_size.width * pattern_size.height;
    cvGoodFeaturesToTrack( image, eig, temp, corners, 
&count,
        quality, min_distance, 
0300.04 );
    cvFindCornerSubPix( image, corners, count,
        cvSize(win_size,win_size), cvSize(
-1,-1),
        cvTermCriteria(CV_TERMCRIT_ITER
|CV_TERMCRIT_EPS,20,0.03));
    
    cvReleaseImage( 
&eig );
    cvReleaseImage( 
&temp );

    
return 1;
}


void  makeChessBoard()
{

  CvScalar e; 
  e.val[
0=255;
  e.val[
1=255;
  e.val[
2=255;
  cvSet(chessBoard_Img,e,
0);
  
for(int i = 0;i<ChessBoardSize.width+1;i++)
      
for(int j = 0;j<ChessBoardSize.height+1;j++)
      
{
          
int w =(image_width)/2/(ChessBoardSize.width);
          
int h = w; //(image_height)/2/(ChessBoardSize.height);

          
int ii = i+1;
          
int iii = ii+1;
          
int jj =j+1;
          
int jjj =jj+1;
          
int s_x = image_width/6;          

        
if((i+j)%2==1)
           cvRectangle( chessBoard_Img, cvPoint(w
*i+s_x,h*j+s_x),cvPoint(w*ii-1+s_x,h*jj-1+s_x), CV_RGB(0,0,0),CV_FILLED, 80 );
      }

}

  • 0
    点赞
  • 4
    收藏
    觉得还不错? 一键收藏
  • 32
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论 32
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值