Bomber19's profile - activity

Hi everybody, i'm using the SURF-Algorithm for some object recognition tasks. In this paper they tested the algorithm with a descriptor length of 36 and the results were good. I've tried to change the OpenCV-Source and rebuild it to test it too, but it doesnt really work. My question is if i just have to change the surf.cpp file and rebuild it or are there other tasks i have to do before rebuilding?? I'm using VS2010.

To use SIFT and SURF now you've got to include the nonfree module. The algorithms were shiftetd to that. - #include <opencv2/nonfree/nonfree.hpp>

Hello, i'm using OpenCV for some objectrecognition application. When i calculate KeyPoints with a method like SIFT the KeyPoints have a parameter size. In the documentation it says that this is the diameter of KeyPoint. My question is how this is calculateted? Or is the parameter depending on the scale where a KeyPoint is found? And if yes, how?

I need this information for a documentation about the project that i've benn working on. And in the paper the descriptorsize is not detailed explained(size in pixel). Also the sourcecode couldn't help me out to get the real size that the descriptorwindow becomes .

So i devide the 16x16 pixel region in 4x4 blocks and calculate a histogramm for each block? Right?

Hi, i'm working with SIFT feature and need some information about the descriptor. In the origanal paper they say that the descriptor use a window of the size 16x16 around a keypoint with 4x4 subregions. My question is if they mean 16x16 pixel? Or is the size(in pixel) of the window depending on the scale? And if yes, how??

Hey, i'm using the RANSAC-Algorithm for estimating homography. The only parameter i can edit is the reprojection error threshold t. What are the values of the other parameters in OpenCV?

• N --> number of trials
• T --> threshold for the number of inliers
• p --> probability that one of the random sample is free from outliers

Hello, i try to get the generic_desciptor_match.cpp file run with the FernDescriptorMatcher. I call it like this:

generic_descriptor_match.exe image.bmp ref_image.bmp FERN params.xml

The problem that i've got is, that it can't create the descriptor and i don't know why. The XML-file with the params looks like this:

• < ?xml version="1.0"?>
• < opencv_storage>
• < nclasses>50</nclasses>
• < patchSize>31</patchsize>
• < signatureSize>2147483647</signaturesize>
• < nstructs>50</nstructs>
• < structSize>9</structsize>
• < nviews>1000</nviews>
• < compressionMethod>0</compressionmethod>
• < /opencv_storage>*

The created descriptorMatcher is always null. Why doesn't it work?

Hi. I'm working with VS2010. So i can build the new OpenCV to another folder and create a new propsheet with the specific path to the include-, lib-files,... in VS. Then it should work?

Hi, i would like to install another OpenCV version (my current is 2.4.2) because SURF is not rotation invariant. Will there be any problems so that after installation of the new one the current version is not working correct? Is there something to consider?

Hi, i'm trying to run the planar object detection algorithm FERN. Here is the code i'm using and my question is if somebody could tell me what is my fault und how i've got to train the Ferns right.

At first i've get the keypoints of the referenceimage with the help of the FAST detector. When this is done i initialize a FernDescriptorMatcher und add the image with the keypoints to it and call the train function. Something wrong?

FernDescriptorMatcher fern;
fern.add(images,keypoints);
fern.train();

After that i load another image with the object i've trained in it and get the keypoints so that i can match them by calling the match function.

fast.detect(testimage,keypoints_testimage);
fern.match(testimage,keypoints_testimage,correspondence);

What i do not understand is why the match function tooks the same time then the train function and there is no feasible result. So how i use the FernDescriptorMatcher correct?

Can somebody tell me how to use it right?

Hi, i try to run the generic_descriptor_matcher.cpp file with Ferns but it always says that is not able to create the descriptor. I've also tryed to get the FernDescriptorMatcher class run by myself by adding some training images and then call the train function but i don't get any feasible result. My Question is how do i use the FernDescriptorMatcher class correctly(Adding Images --> Training ---> Matching)?

Hey, do you know any paper where it is explained more specific?

Thanks ;)

I've already transformed the images but have no idea how i can figure out the change of the pixelposition. Normally i'm working with Matlab. There is a function is available for this. Do you have any idea how to do this?

Hello, I've got the following problem: I would like to transform an image (with scaling, rotation and sheering) and save the position of a pixel in the reference image and the transformed position of this pixel in the transformed image. How can I do this, when the transformation matrix is known and I work with Mat objects?

Thanks for the answer. I've got 90 detected points on a 320x240 image. I've figured out that it takes a lot of time to train because of my computer. But now i've got a new question. Why does the FernDescriptorMatcher call the train() function again when i call the match() function. And how can i change that?

Hey! I'm trying to train a FernDescriptorMatcher. First i extract keypoints with SURF und then i call the train method with only one image(for test). I've been waiting several minutes but the training doesent end.

Why?

Hi, i'm trying to train a Fern-Classifier for feature matching. My problem is that i don't understand how it works. I've tryed to understand the OpenCV samples and also studyed the documentation but cant solve the problem. Can somebody tell me how it works or give me some tips? I'm also interested in other methods for robust real-time feature matching.