A corner can be located by following these steps:
- Calculate the gradient for a small window of the image, using sobel-x and sobel-y operators (without applying binary thesholding).
- Use vector addition to calculate the magnitude and direction of the total gradient from these two values.
 |
|---|
| vector addition |
- Apply this calculation as you slide the window across the image, calculating the gradient of each window. When a big variation in the direction & magnitude of the gradient has been detected - a corner has been found!
for corner detection opencv's (cornerHarris()) feature is used with parameters being [float image , window size , sobel operators size,constant]
for documnetation you can refer here