rtlumb's profile - activity

Found answer, but it won't let me answer my own question as a new user. Answer: http://www.answers.opencv.org/questio...

Also the previously asked questions on this topic that I searched for are left remarkably unresolved. Would be nice to answer this definitively and have a resource for others later on.

Hello, I'm having problems locating the convention for the origin of the world coordinate system for a stereo camera pair in the opencv doc. I believe the origin is centered on the left most camera and along the left most camera's field of view as I believe this is typical convention. I'm currently having a strange, unexpected shift in the x-coordinate of my outputted world coordinates of an LED target. In the past this has been a calibration error, but if someone could provide the openCV standard for the origin point in a stereo vision system with a link to doc, that would be very useful.

Thanks.

The proper intensity value was able to be accessed with the .at() function once the proper object types were used. In this case, and unsigned char object was being assigned to a char object.

"The typical effect (for almost every C++ compiler you'll use) is that 0-128 is the same, but anything above that becomes negative." -Tetragramm

Once the conversion was from uchar directly to int, instead of uchar to char to int, the expected 0-255 range of intensities were observed.

unsigned char g = image.at<uchar>(center); int r = int(g); cout << "r: " << r << endl;

Special Thanks to Tetragramm for helping resolve the issue.

-Rowan

I have tried that with code like:

char g = image.at<uchar>(center); int r = int(g); cout << "r: " << r << endl;

Unfortunately this gives me the same output as the printf line above, and does return an integer value, but nothing like I'd expect. The values I currently receive are single digits values while I expect something about 150 or above.

I have tried switching the rows and cols to no avail. I think the problem is a bit deeper than that. Also, shortened the title a bit.

In a larger context, I'd like to perform a sweep on detected circle object in a image, but cannot get a reasonable intensity value or I guess access the pixel correctly. I expect a uchar pixel type with values from 0-255. When I run the code, I get outputs that either involve letters or numbers that are much too low to be correct.

I use cvtColor() to first change a RGB image to Gray. The RGB image is originally a CV_8UC3 type mat image. I am also using openCV 3.0.

Here's a small snippet of code where I try to display a pixel value in the terminal. Here, circles is a vector of vectors, and I'm using it to access the center point of the detected circle object. When I draw the detected circles on the image and display it, there seems to be no issue.

Point2f center(cvRound(circles[0][0]), cvRound(circles[0][1]));
int x = round(circles[0][0]);
int y = round(circles[0][1]);
printf("center: %x\n", image.at<uchar>(center));
printf("center: %x\n", image.at<uchar>(circles[0][0], circles[0][1]));
printf("center: %x\n", *image.ptr<uchar>(x, y));

Thanks! Rowan