RonM's profile - activity

Hi all, I downloaded opencv 2.4.9 and managed to get Python2.7 up and running with opencv. But when I checked the version using

import cv
cv2.__version__

I got

$Rev: 4557 $'

From what I read on other threads, this is an old version and I can also see that the interpreter is using opencv-2.3.1 for some reason. So I don't have FAST,SURF and all those goodies.

I used the following build command from this site https://www.raben.com/book/export/html/3

cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local
-D WITH_TBB=ON -D BUILD_NEW_PYTHON_SUPPORT=ON -D WITH_V4L=ON 
-D INSTALL_C_EXAMPLES=ON -D INSTALL_PYTHON_EXAMPLES=ON 
-D BUILD_EXAMPLES=ON -D WITH_QT=ON -D WITH_OPENGL=ON ..

This question is very similar to my problem http://stackoverflow.com/questions/18561910/opencv-python-cant-use-surf-sift

But I don't have the old 2.4.6 installation suggested and would also like to use a newer version if possible.

Already spent a good bit of time trying to fix this but with no success. Any help would be much appreciated!

Thanks, Ronen

Hi,

I'm trying to use the opencv stitching pipeline to develop a prototype for a custom stitching flow. The stitching is between 2 pictures, the resolution of each is about 1200x960.

I noticed that the warp perspective function is pretty lengthy, roughly 0.5 seconds for each picture.

Any estimates of how much this could be improved upon and by what method (besides improving hardware)?

Thanks in advance!

Ron

Have you tried using the stitching module for the setup? I've used it successfully for a setup with 2 cameras 5 cm apart (not a single rotating camera) .

Hi Balaji, I don't have photos, I need to create the setting which I will then photograph. I think I wasn't clear enough as to what I'm trying to do, so I updated the original post to clarify my goal.

Hi,

I would like to stitch two photos (horizontally) with a minimum of overlap between them. But I can control what is in the overlap of the two photos, as indicated by the orange segment shown in the picture.

For example, I could print out a pattern that will help the feature finding step in the open-cv stitcher algorithm. Using the SURF feature detector.

My question is: what sort of pattern (i.e chessboard, or something with lots of corners, etc) would be the best for the task of feature finding and stitching the photos? Best = most pairwise matches can be found between the two photos.

I will clarify my goal: I have two cameras who are fixed one relative to the other (assuming negligible distance between them for homography calculation purposes) and I would like to use the stitching process to find the transformation between the cameras. So I don't have actual photos, my aim is to create a "calibration target" which I can photograph using the 2 cameras. So I can decide what the photos will be of, and don't have any 2 specific photos right now. I want a minimum overlap between the fields of view of the cameras (not more than a sheet of paper at 10 meters distance from the cameras), that's why I'm trying to make what little overlap I have count the most (thus I want to "pack" it with easy to identify features). Then, after I have computed the transformation from the calibration phase, I can use it to speed up the stitching process for images taken from the cameras in the future.

Thanks in advance,

Ron

Hi all, I'm new to CV, and trying to stitch together a video of two cameras which are stationary one relative to the other. The details:

The cameras are one beside the other and I can adjust the rotation angle between them. The cameras will be moving with respect to the world, so the scene will be changing.

The amount of frames to be stitched is roughly 300 (each frame is composed of two pictures, one from each camera).

I don't need to do the stitching in real time, but I want to do it as fast as possible using the fact that I know the relative positions of the cameras. Resolution of each picture is relatively high, around 900x600.

Right now I'm at the stage where I have code to stitch 2 single pictures, courtesy of http://ramsrigoutham.com/2012/11/22/panorama-image-stitching-in-opencv/

The main stages are:

1. Using SURF detector to find SURF descriptor in both images
2. matching the SURF descriptor using FLANN Matcher
3. Postprocessing matches to find good matches
4. Using RANSAC to estimate the Homography matrix using the matched SURF descriptors
5. Warping the images based on the homography matrix

My question is: How can I optimize the process based on the fact that I already know the camera positions?

Ideally I would like to do some initial calculation once to find the transform between the camera perspectives, and then reuse it. But not sure with my rudimentary CV knowledge if this is indeed possible, and what transform I could use if so.

I understand that calculating the homography matrix once and reusing it won't work, since the scene is changing.

Two other possibilities:

1. I found a similar case (but stationary scene) where the transform is computed once and reused. Which transform is this, and could it work in my case?

2. The other possibility I found is to use the initial knowledge to find the overlapping region between two pictures, and ignore the rest of the pictures to save time. Relevant thread

Any help would be greatly appreciated!
Ron