TLDR; Your model is the mathematical description, for example, just using an intrinsic matrix without distortion: K = [fx, 0, cy; 0 fy, cy, 0 0 1] and your calibration for this model and the actual camera model you are using could be: fx = fy = 105.2 (focal length in terms of pixels), cx = 961.8, cy = 539.8. You're probably not going to ever read this, but for anyone interested: The camera model is the mathematical description of the camera. Typically you would have at least a K matrix (3x3) which includes the focal length and optical center. Commonly, however, you would in-cooperate at least some way of dealing with lens distortion. Typical ways of modeling distortion rely on making the optical center a function of position on the image sensor. Specifically, as a function of the distance to the optical sensor, such that pin cushion distortion and barrel distortion can be modeled. Of course, you can make this more complex, but this will also make calibration more difficult. So what is calibration? Calibration is figuring out the values of the parameters to your camera model, such that when projecting world points P to image points p onto your sensor (say p = K * M * P, with M being the extrinsic matrix [R t]) you minimize the error between observed points and projected points. Once the estimated values of your model (your calibration) match close enough to the "real" values, your projected points will be accurately projected to within a pixel (typically below 0.3 is easily attainable).
![Blox Fruits Dragon Rework Update [Full Stream]](http://i.ytimg.com/vi/EqDAp8udhm0/mqdefault.jpg)
I can't thank you enough. You are really a good lecturer. Learned everything about camera model and calibration lecture from you. Thank you so so much
Was struggling for hours, Thank god for this video!!!
One thing remains a mystery: what are the values for a31.....a34 in the Matrix A?
Anyone know if Ch3 is not used, why the camera model matrix has to be 4x4, instead of using 3x4 matrix?
Thanks for the lectures. At 34.28 third component of homogeneous co-ordinate made zero. this give one more point on the same line. How ?
Thank you you've helped me a lot!
I have one query that is whats the difference between camera model and camera calibration
TLDR; Your model is the mathematical description, for example, just using an intrinsic matrix without distortion: K = [fx, 0, cy; 0 fy, cy, 0 0 1] and your calibration for this model and the actual camera model you are using could be: fx = fy = 105.2 (focal length in terms of pixels), cx = 961.8, cy = 539.8.
You're probably not going to ever read this, but for anyone interested: The camera model is the mathematical description of the camera. Typically you would have at least a K matrix (3x3) which includes the focal length and optical center. Commonly, however, you would in-cooperate at least some way of dealing with lens distortion. Typical ways of modeling distortion rely on making the optical center a function of position on the image sensor. Specifically, as a function of the distance to the optical sensor, such that pin cushion distortion and barrel distortion can be modeled. Of course, you can make this more complex, but this will also make calibration more difficult.
So what is calibration? Calibration is figuring out the values of the parameters to your camera model, such that when projecting world points P to image points p onto your sensor (say p = K * M * P, with M being the extrinsic matrix [R t]) you minimize the error between observed points and projected points.
Once the estimated values of your model (your calibration) match close enough to the "real" values, your projected points will be accurately projected to within a pixel (typically below 0.3 is easily attainable).
Calibration, without "e".
Thank you for the lecture.
just realize that the persective matrix does not work when Z is zero.
all of you who talk about photogrammetry outside the west , do not understand the mathematical models as it is ! (