I'm doing a project where I'm extracting keypoints using the AGAST algorithm. One curious thing popped out to me - keypoint.size for AGAST keypoints evaluates to 7.0 for every keypoint. Why is this? Is there something I'm missing with the AGAST algorithm?
I think (may be i am wrong) that there is no meaning. If you read source code 7.0 is defined here or hereor https://github.com/opencv/opencv/blob.... it's always 7.0. keypoint constructor needs a sizewhen you give a location. Why 7.0 ? ask to author (https://github.com/cbalint13?) may be read this https://en.wikipedia.org/wiki/7
Yeah that makes sense. But isn't it the size of a keypoint that is used when computing the overlap error for computing e.g. reliability though? Or have I misunderstood?
you are right so 7 is not here may be read this https://en.wikipedia.org/wiki/7 but here : may be read this https://mediatum.ub.tum.de/doc/128745... : Like SUSAN, also FAST uses a Bresenham’s circle of diameter 3.4 pixels as test mask (Adaptive and Generic Corner Detection Based on the Accelerated Segment Test) 3.4 * 2=6.8 read section 4.1 too
Yeah it makes sense that it comes from the radius of the Bresenham's circle that is used in order to evaulate the interest point. So, a correspondence between two AGAST points is going to be very rare, isn't that right? Considering that the keypoint region is so tiny I mean.
Also, I suspect you didn't intend to link to the wikipedia article on the number 7. :D
Asked: 2019-02-07 07:21:55 -0600
Seen: 242 times
Last updated: Feb 07 '19
