Revision history [back]

Hello Forum,

I am trying this tutorial out http://www.learnopencv.com/blob-detection-using-opencv-python-c/ and I have a question. How do i keep the best detected blob in an image based on its "response". The parameters for KeyPoint http://docs.opencv.org/master/d2/d29/classcv_1_1KeyPoint.html has a response.

I have been trying and i keep getting a 0 (because the default argument is 0).

Thanks

Hello Forum,

I am trying this tutorial out http://www.learnopencv.com/blob-detection-using-opencv-python-c/ and I have a question. How do i keep the best detected blob in an image based on its "response". The parameters for KeyPoint http://docs.opencv.org/master/d2/d29/classcv_1_1KeyPoint.html has a response.

I have been trying and i keep getting a 0 (because the default argument is 0).

ThanksThanks!

EDIT :

This is what I have been trying to do. Detect the blobs in the image below and find out how good the fit is. Input is Figure 1.

In Figure 2, the output of the blob detector is shown. What is missing is the "goodness of fit". In Figure 3, I went to find the contours and then computed the circularity. It is still not too bad.

The problem would be in Figure 4 (paint drawing). Finding contours and computing the circularity would give me very bad results. I would like to be able to detect a blob which will encompass ONLY the semicircle. Of course there may be other blobs within the white regions but I don't care.

Basically I would like to find possible locations in an image to fit a circle/blob.

Possible solutions

I have been reading and I found out about the star feature detector in opencv. Would it work and will I get the "strength" of the match ? I also read about haar pointed features. http://athenanichol.com/blog/?p=522 Would those work ? Please help, thanks.

Figure 1

Figure 2

Figure 3

Figure 4

Hello Forum,

I am trying this tutorial out http://www.learnopencv.com/blob-detection-using-opencv-python-c/ and I have a question. How do i keep the best detected blob in an image based on its "response". The parameters for KeyPoint http://docs.opencv.org/master/d2/d29/classcv_1_1KeyPoint.html has a response.

I have been trying and i keep getting a 0 (because the default argument is 0).

Thanks!

EDIT :

This is what I have been trying to do. Detect the blobs in the image below and find out how good the fit is. Input is Figure 1.

In Figure 2, the output of the blob detector is shown. What is missing is the "goodness of fit". that there is no parameter telling me how good the fit is. In Figure 3, I went to find the contours and then computed the circularity. It is still not too bad.

The problem would be in Figure 4 (paint drawing). Finding contours and computing the circularity would give me very bad results. I would like to be able to detect a blob which will encompass ONLY the semicircle. Of course there may be other blobs within the white regions but I don't care.

Basically I would like to find possible locations in an image to fit a circle/blob.

Possible solutions

I have been reading and I found out about the star feature detector in opencv. Would it work and will I get the "strength" of the match ? I also read about haar pointed features. http://athenanichol.com/blog/?p=522 Would those work ? Please help, thanks.

Figure 1

Figure 2

Figure 3

Figure 4