Prev: Joint Hypothesis
Next: Warning: No display specified. You will not be able to displaygraphics on the screen.
From: Emre on 21 May 2010 14:08
Hi,
I'm working on a project to detect ships at horizon.
But, first I have to detect horizon and then work on object above the horizon.
I'd be glad to receive any help.
From: Mark Shore on 21 May 2010 14:18
"Emre " <jasonborn(a)windowslive.com> wrote in message <ht6i65$bko$1(a)fred.mathworks.com>...
> Hi,
> I'm working on a project to detect ships at horizon.
> But, first I have to detect horizon and then work on object above the horizon.
> I'd be glad to receive any help.

Google it, or work it out from scratch. Only simple trigonometry is required and it will be trivial to implement in MATLAB.
From: Doug Hull on 21 May 2010 14:19
"Emre " <jasonborn(a)windowslive.com> wrote in message <ht6i65$bko$1(a)fred.mathworks.com>...
> Hi,
> I'm working on a project to detect ships at horizon.
> But, first I have to detect horizon and then work on object above the horizon.
> I'd be glad to receive any help.

What kind of sensors? Do you have typical (assuming) images with and without ships present? What have you tried?
From: ImageAnalyst on 21 May 2010 14:54
On May 21, 2:08 pm, "Emre " <jasonb...(a)windowslive.com> wrote:
> Hi,
> I'm working on a project to detect ships at horizon.
> But, first I have to detect horizon and then work on object above the horizon.
> I'd be glad to receive any help.

----------------------------------------------------------------
Post some images. Presumably this is just a student project because
in these days of radios, telephones, and GPS, I don't see any
practical use for this. But this means that you'll probably have nice
photos with ships on the horizon. I've lived on the California coast
so I know that often (maybe most of the time?) you don't see the
horizon. I know that seeing a sun sink below the horizon is a fairly
scarce event - most of the time the sun sinks into the "marine
layer" (which is a layer of fog above the ocean) and fades away, much
to the displeasure of visiting tourists, although you can see the
horizon sometimes during the middle of the day when the marine layer
has retracted from the coast. But you'll probably have a set of nice,
perfect sample images with clearly distinct sky, water, and ship,
where you just have to find the top rectangular region and the bottom
rectangular region, using texture, color, etc. and then find the
horizon line. Maybe you'll use edge filters also, or maybe you'll use
a combination of things.
Good luck.
From: Emre on 22 May 2010 11:16
ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <ec52bdcb-45e0-4ed5-bbe8-4267a4f43558(a)e28g2000vbd.googlegroups.com>...
> On May 21, 2:08 pm, "Emre " <jasonb...(a)windowslive.com> wrote:
> > Hi,
> > I'm working on a project to detect ships at horizon.
> > But, first I have to detect horizon and then work on object above the horizon.
> > I'd be glad to receive any help.
>
> ----------------------------------------------------------------
> Post some images. Presumably this is just a student project because
> in these days of radios, telephones, and GPS, I don't see any
> practical use for this. But this means that you'll probably have nice
> photos with ships on the horizon. I've lived on the California coast
> so I know that often (maybe most of the time?) you don't see the ...




First of all I tried to use "canny edge detection" to find edges. Then I used "hough" trasnsform to find the longest horizontal line. And I used this line to draw horizon line.

Then, I used medfilt2 on original image to remove noise.
Next, I used canny edge detection to detect edges
then, I applied conv2 to reduce the number of connected components
after that, I found the connected components with "bwlabel"
then, I used "regionprops" to measure properties of image regions.
Then I wanted to eliminate the parts that have below the area of 500 pixel
finally, I wanted to get the regions match to this criteria in "boundingBox"
When I erase the pixels near the horizon when not erasing this pixels multiple regions perceive as one connected component because of the horizon line..(But, I'm not sure that erasing pixels below horizon is the correct approach).
When there is nothing except the ship or land above horizon I can take them in a box
But, when there is land behind the ship I can't discriminate the ship and binary image gets too complex. Even if I can find an object on the surface how can I know whether it's a ship or land.
I will be very grateful if you or anyone can help.(I'm in a great stress because of the possibility of not being able to finish the project on time)
Here is the source code I've written so far. The links for the pictures are below the source code.
%detection.m
A = imread('im2.jpg');
R = imresize(A, [333 500]);
I = rgb2gray(R);
BW = edge(I, 'canny', 0.1);
%hough transform to find the lines
[H, T, R] = hough(BW);
P = houghpeaks(H, 5, 'threshold', ceil(0.3*max(H(:))));
x = T(P(:,2));
y = R(P(:,1));
lines = houghlines(BW,T,R,P,'FillGap',4,'MinLength',7);
figure, imshow(I), hold on
max_len = 0;
for k=1:length(lines)
xy = [lines(k).point1; lines(k).point2];
%Find the length of each line
len = norm(lines(k).point1 - lines(k).point2);
if (len > max_len)
%change the value of max_len when finding a
%longer line
max_len = len;
%find the longest line at most that has 0.1 degree
%because I thought the horizon should be straight
if (atan((xy(2,2)-xy(1,2))/(xy(2,1)-xy(1,1)))<0.1...
&& atan((xy(2,2)-xy(1,2))/(xy(2,1)-xy(1,1)))...
> -0.1 && (xy(2,2)>150))
xy_long = xy;
end
end
end
%draw the longest line from one side of the picture
%to other side
xy_long(1,1) = 0;
xy_long(2,1) = 500;
plot(xy_long(:,1),xy_long(:,2),'LineWidth',2,'Color','cyan');
%------------------------------------
%Median filtering the image to remove noise
im1=medfilt2(I,[3 3]);
BW1 = edge(im1,'canny',0.4); %finding edges
[imx,imy]=size(BW1);
msk=[0 0 0 0 0;
1 1 1 1 1;
1 1 1 1 1;
1 1 1 1 1;
0 0 0 0 0;];
%Smoothing image to reduce the number of connected components
BW1=conv2(double(BW1),double(msk));
%erase the pixel values near horizon
%before finding connected components
y_begin = xy_long(1,2)-5;
y_end = xy_long(1,2)+5;
for k = y_begin:y_end
for m = 1 : 500
BW1(k,m) = 0;
end
end
L = bwlabel(BW1,8);% Calculating connected components
%I wanted to find the region properties
stats = regionprops(L,BW1,{'Area','BoundingBox'});
%find areas more than 500 pixels
idx = find([stats.Area]>400);
for k=1:length(idx)
%find the regions above horizon
if(stats(idx(k)).BoundingBox(2) < xy_long(1,2))
rectangle('Position',stats(idx(k)).BoundingBox,...
'EdgeColor','y');
end
end
hold off
figure, imshow(BW1)
%*********************
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im10.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im8.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im2.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im7.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im12.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im5.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im9.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im6.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im1.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im4.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im3.jpg
http://i804.photobucket.com/albums/yy326/lifesgood_photos/im11.jpg
 |  Next  |  Last
Pages: 1 2
Prev: Joint Hypothesis
Next: Warning: No display specified. You will not be able to displaygraphics on the screen.