Opencv_cvfindcornersubpix () Finding sub-pixel corner points

If the purpose of our image processing is not to identify feature points but to perform audit measurements, we usually need more precision, whereas cvgoodfeaturetotrack () can only provide simple pixel coordinate values, but sometimes we need actual coordinate values instead of the certificate coordinate values, for example, We want to determine the position of a sharp peak point in the graph, but the position of the peak point is generally not at the center of a pixel, and subpixel detection can be used.

The position of the sub-pixel corner is a basic measure when the star is calibrating, tracking and reconstructing the Serpentine's trajectory or reconstructing the three-dimensional structure of the tracked target. The Cvgoodfeaturestotrack () function can be used to obtain the coordinates of a corner point, and the next step is to discuss how the obtained coordinate values are accurate to sub-pixel accuracy. The method is the dot product theory of vectors: The dot product of the uoge vector and its orthogonal vector is 0, and the corner point we have previously said, is the intersection of two edges, can satisfy this situation. Such as:

When the required point P is in the interior of an area, the point P edge is flat and its gradient value is 0, when the vector

voidconst CvArr* image, CvPoint2D32f* corners,int count, CvSize win, CvSize zero_zone,CvTermCriteria criteria );

Image
Enter an image.
Corners
Enter the initial coordinates of the corner point, and also store the exact output coordinates
Count
Number of corner points
Win
Half the size of the search window. If win= (5,5) then use 5*2+1x5*2+1 = 11x11 size of the search window
Zero_zone
The dead zone is half the size, and the dead area is the area where the central location of the search area is not summed. It is used to avoid some possible singularity of the autocorrelation matrix. A value of ( -1,-1) indicates no dead zone.
Criteria
The termination condition for the iterative process of the angular point. That is, the position of the corner is determined, either the number of iterations is greater than a certain value, or the accuracy reaches a certain setpoint. Criteria can be the maximum number of iterations, or a set of precision, or a combination of them.
The function Cvfindcornersubpix through iteration to find the position of a corner point with sub-pixel precision, or the radial saddle point (radial saddle points).

When the party Zhao comes to a new position in the Q, the algorithm iterates over the new corner as the initial point to know that the iteration termination condition satisfies the user definition.

Opencv_cvfindcornersubpix () Finding sub-pixel corner points