sachin_rt's profile - activity

I have seen that but didn't understand it If a pixel is not too close to neutral or its color is well defined , How should my S min , V min be to ignore such a color ?

Can anybody help me understand the Vmin , Vmax and S min thresholds in the camshift Demo ?

How is the Hue channel more advantageous than others for object tracking ?

Is the coord.x in the " KF.statePre.at<float>(0) = coord.x; " and in the " this->measurement(0) = coord.x; " same ??

Heyy thanks for that quick reply :) I am trying to track a face using a PTZ Camera . For now I am doing face detection using Viola Jones , then I am tracking the detected face using camshift tracking and moving the camera accordingly . But this works a bit slowly . I want to make this more robust and faster , Can you help me with any ideas for faster tracking ??

I am detecting the face using Haar Detection , But the detection is very slow and not very accurate ? So I was thinking if I could use kalman filter for face detction things would be much better . Any body who worked on this before please help ?

Edit : Is there any other way I can make the face detection robust ? I am working on real time video capture .

Edit : I am using a PTZ camera to track the face and move the camera accordingly so as to maintain the face in the center region . Are there ways we can predict the position of the face ?? Note that here the camera has motion , ie pan and tilt . So what are the ways we can predict the future position of the face ??

Are fx and fy not the physical focal length values ?? If they are then fx and fy have to be constant right ??

Hii I am using a logitech PTZ camera . i am trying to find the intrinsic camera matrix from opencv . Do the fx and fy values change when calibrated differently ?? Or are they constant ?

Hii I want to find the distance of an object from the camera ? Camera has got autoFocus Property . Is it possible to find the focus distance in opencv ??

Hii I want to get real world (X,Y,Z) coordinates of an object from live capture from a PTZ camera . I have found the intrinsic parameters , and using chessboard calibration with 15 chessboard images . I have also found the extrinsic prameters . I know Z coordinate cannot be found using a single camera .

To do this I need to find rotation and translation matrix . I have a doubt that for finding rotation matrix cvRodrigues() function will convert a rotation vector to 3x3 rotation matrix , but here I will have 15 rotation vectors . Which one should I use for finding the rotation matrix ??

Also I want to know if I happen to pan or tilt my camera from the original calibration position . will i have to recalculate my rotation matrix and translation matrix or can i use the old ones ??

I am using opencv 2.4.9 .I am capturing live from a webcam and want to extract properties like frame rate , frame count , timestamp , etc . I heard that the videoInput library will be helpful in this case and learnt that it has been integrated into opencv .Can anybody give me a small piece of code how I can use the videoInput functions in opencv . I have tried in this fashion , but have been unsuccessful.

CVCapture* capture = CVCaptureFromCAM(CV_CAP_DSHOW)

IplImage * image = cvQueryFrame(capture);

opencv mentions that CV_CAP_DSHOW means we can use DirectShow via VideoInput . Now how do I use the functions of videoInput for ex , listOfDevices() of videoInput which will give me the number of devices attached ?? What are the extra advantages I get from mentioning CV_CAP_DSHOW and how do I use them ?

Hii I want to convert pixel coordinates from an image to real world 3D coordinates . I have used camera calibration using a chessboard and have found Rotation Matrix , Translation matrix , Intrinsic matrix . What should be my next step for conversion ?? Essentially I want the 3D projection of a pixel of an image . Is it possible ?? Please Explain

When I am trying to get the frame number and Milli sec time using CvGetCaptureProperty() I am getting -1 all the time ?? I have searched a lot on the net but haven't found any solutions . Please help if anybdy has got it solved Brief Code looks like this :

CvCapture* capture = cvCaptureFromCAM(0);

while(1) { IplImage* frame = cvQueryFrame(capture);

cout << (int) cvGetCaptureProperty(capture,CV_CAP_PROP_POS_MSEC);

cvShowImage("Video_Input",frame);

if ( (cvWaitKey(10) & 255) == 27 ) break;

}

Hii , I am using cvCalibrateCamera2(....) function in opencv . Here one of the output that I get is the rotation_vectors which is Nx3 matrix . I have seen the documentation and it says look at cvRodrigues2() function for further details . And I have understood that cvRodrigues2() function converts the a 1x3 rotation vector to a 3x3 rotation matrix . My question is which 1x3 rotation vector out of the N , should be inputted to cvRodrigues2() function for calculating the Rotation Matrix ??

Heyy bjoernz thanx fr the quick reply :) I have one more query . What is the reference coordinate system ? We have formulae where we substitute the pixel coordinates and get the world coordinates , What is the reference coordinate system ??

Is it possible to get 3D coordinates of the objects with a single camera ??

What does HomographyMatrix*HomogeneousPixelCoordinates give ? I think the result would be world coordinates right ?? http://stackoverflow.com/questions/7836134/get-3d-coord-from-2d-image-pixel-if-we-know-extrinsic-and-intrinsic-parameters/10750648#10750648