/tag_detector

Answer for Bito Company

Primary LanguageMakefile

tag_detector

Answer for Bito Company by yp

1、STEPS

STEP 1 : Compile project

    $cd tag_detector/src
    $catkin_make

STEP 2 : launch the node for publishing video

    $roscore
type in a new terminal
    $source devel/setup.bash
    $rosrun tag_detector video_frame_pub

STEP 3 : launch the node for subscribing video

type in a new terminal
    $source devel/setup.bash
    $rosrun tag_detector video_frame_sub

STEP 4 : call rosservice

type in a new terminal
    $source devel/setup.bash
    $rosservice call set_blur_window_size 11 11 
        (the window size should be odd and > 0 )

2、Result

The relationship between nodes:

avatar

The result(one frame of all images):

avatar

3、problems

In the question 8 , i think the question should be solved like this:

J = I o K；

I represents original image; o represents convolution; K represents blur kernel; J represents the blurred image;

So we should implement a deblur filter like this:

FFT(J) = FFT(I o K) = FFT(I) × FFT(K);
I = IFFT( FFT(J) / FFT(K) );

And I take the wiener filter and Richardson_Lucy Algorithm as a comparison

avatar

The right upper image is the blurred image with Gaussian kernel.

The right lower image named Deconv image was got by wiener filter ,somehow the content of image was damaged. Because of time , i give up the method.

The left lower image was got by Richardson_Lucy Algorithm.However,it needs many iterations to improve the result, so applying it to the video deblurr problem is not a good idea!

So my result was not as good as i imagined because of lack of time.