Shulou(Shulou.com)06/01 Report--

What this article shares with you is about how FASTN can quickly detect corners. The editor thinks it is very practical, so I share it with you to learn. I hope you can get something after reading this article.

Draw a circle with a radius of r, covering the M pixels around the p point.

Set the threshold t, if there are 16 consecutive pixels around the pixel value minus Ip − t, or if there are consecutive N pixels greater than Ip+t, if t is 0, then it can be understood that there are consecutive N pixels greater than or less than the gray value of Ip. Then this point is judged to be the corner.

A faster improvement is to first detect whether three of the four points around p-point, that is, 1,5,9,12, satisfy more than Ip+t, if not, skip it directly, and if so, continue to use the previous algorithm to determine whether 12 of the 16 points meet the conditions.

The disadvantage of the above algorithm is that most of the detected points are adjacent to each other, and we need to remove some of these points. In order to solve this problem, the algorithm of non-maximum suppression can be adopted: assuming that the two points of P _ (Q) Q are adjacent, the difference and V between the two points and the 16 pixels around them are calculated respectively, and the points with small V values are removed, that is, the non-maximum corners are suppressed.

So the question is, can corners of any angle be detected? As shown in the following figure: there are three kinds of corners, namely, 45 °, 90 °and 135 °.

So can any corner be detected by FASTN algorithm?

The answer is yes. But it depends on the setting of consecutive pixels N. We still say that if the radius is 3, then there are 16 pixels on the circle: as shown below:

Because the condition of the algorithm to detect corners is that the continuous N pixels are greater than or less than the central gray value minus the threshold t, so this N determines the angle that can be detected to some extent. For example, the following figure: with the black line as the benchmark, the angle less than 135 °(black-red angle) can be detected if you set Numb10, the angle less than 90 °(black-blue angle) can be detected if you set Numb12, and the angle less than 45 °(black-green angle) can be detected if you set Numb14.

3. FASTN algorithm

1) the speed is much faster than other algorithms.

2) it is greatly affected by the image noise and the threshold set.

3) FASTN does not produce multi-scale features and FASTN feature points have no direction information, so rotation invariance will be lost.

This is how FASTN can quickly detect corners. The editor believes that there are some knowledge points that we may see or use in our daily work. I hope you can learn more from this article. For more details, please follow the industry information channel.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.