MATLAB代做|FPGA代做--基于MATLAB的EMD算法的仿真程序

资源简介：

MATLAB程序下载——一基于MATLAB的EMD算法的仿真程序。

function [v2, sigma, p, hpos] = getVp(lines, imsize,DO_DISPLAY)

% [v2, sigma, p, hpos] = getVp(lines, imsize, DO_DISPLAY)

% Estimates the principal vanishing points, as in Video Compass [Kosecka

% 2002], except that more (or fewer) than 3 principal vps could be found

% 

% Input:

% lines([x1 x2 y1 y2 angle r])

% imsize: size of image

% DO_DISPLAY (optional): whether to create display figures (default=0)

% Output:

% v2(nvp, [x y]) - the found vanishing points; last is outliers; vp are in

%                  units of pixels/imsize(1), with image upper-left at (0,0)                  

% sigma(nvp) - the variance for each vp

% p(nlines, nvp) - the confidence that each line belongs to each vp

% hpos - horizon position (0 is top of image)

if ~exist('DO_DISPLAY')

    DO_DISPLAY = 0;

end

nlines = size(lines, 1);

x1 = [lines(:, [1 3]) ones(size(lines, 1), 1)];

x2 = [lines(:, [2 4]) ones(size(lines, 1), 1)];

% 锟斤拷锟斤拷叨味锟接ζ斤拷锟侥凤拷锟斤拷

l = cross(x1, x2);

l = l ./ repmat(sqrt(sum(l.^2,2)), 1, 3);

nbins = 60;

theta = mod(lines(:,5), pi);

% 锟斤拷锟絫heta锟斤拷统锟斤拷直锟斤拷图

binwidth = pi/nbins;

bincenters = [(binwidth/2):binwidth:(pi-binwidth/2)];

hist_theta = hist(theta, bincenters);

% 平锟斤拷直锟斤拷图

for b = 1:nbins

    hist_theta(b) = sum(hist_theta(mod([(b-1):(b+1)]-1, nbins)+1) .* [0.25 0.5 0.25]);

end

% 锟斤拷锟斤拷直锟斤拷图锟斤拷C

s = 9;

C = zeros(1, nbins);

for b = 1:nbins

    C(b) = hist_theta(b) - mean(hist_theta(mod([(b-(s-1)/2):(b+(s-1)/2)]-1, nbins)+1));

end

% 寻锟揭凤拷值

zc_pos = find((C > 0) & ([C(end) C(1:end-1)]<0));

zc_neg = find((C < 0) & ([C(end) C(1:end-1)]>0));

ngroups = length(zc_pos);

bc = bincenters + pi/nbins/2;

if zc_neg(1) < zc_pos(1)

    i1 = round((zc_pos(end)+zc_neg(end))/2);

    i2 = round((zc_neg(1) + zc_pos(1))/2);

    groups{1} = find((theta > bc(i1)) | (theta < bc(i2)));

    for i = 2:ngroups

        i1 = round((zc_pos(i-1)+zc_neg(i-1))/2);

        i2 = round((zc_neg(i) + zc_pos(i))/2);     

        groups{i} = find((theta > bc(i1)) & (theta < bc(i2)));

    end

else

    for i = 1:ngroups

        i1 = ceil((zc_pos(i)+zc_neg(max(i-1,1)))/2);

        i2 = ceil((zc_neg(i) + zc_pos(min(i+1,ngroups)))/2);

        groups{i} = find((theta > bc(i1)) & (theta < bc(i2)));       

    end

end

remove = [];

thresh = max(0.05*nlines, 5);

for i = 1:ngroups

    if length(groups{i}) < thresh

        remove(end+1) = i;

    end

end

groups(remove) = [];

ngroups = length(groups);

% 锟斤拷始锟斤拷EM

sigma = ones(1, ngroups); 

p = zeros(nlines, ngroups); 

for i = 1:ngroups

    p(groups{i}, i) = 1;

end

A = l;

v = [];

sigma = [];

for i = 1:ngroups

    normp = p(:, i) / sum(p(:, i));

    W = diag(normp);

    [eigV, lambda] = eig(A'*W'*W*A);

    [tmp, smallest] = min(diag(lambda));

    v(1:3,i) = eigV(:, smallest);

    sp = sort(normp, 'descend');

    sp = sum(sp(1:min(length(sp), 2)));

    sigma(i) = normp' * (l*v(:,i)).^2 / (1-sum(sp));

end

% 锟斤拷b锟斤拷锟斤拷锟斤拷

nadd = 3;

ngroups = ngroups + nadd;

sigma(end+(1:nadd)) = 0.2;

tmpv = [0 0 1]';

v(1:3, end+1) = tmpv / sqrt(sum(tmpv.^2));

tmpv = [1 0 1]';

v(1:3, end+1) = tmpv / sqrt(sum(tmpv.^2));

tmpv = [-1 0 1]';

v(1:3, end+1) = tmpv / sqrt(sum(tmpv.^2));

p(:, end+(1:nadd)) = 0;

pv = ones(1, ngroups);

oldv = v;

% 锟斤拷始EM

for iter = 1:15

    oldp = p;

    pv = pv / sum(pv);

    S = repmat(sigma, nlines, 1) + 1E-10;

    plv = exp(-(l*v).^2 ./ S / 2) ./ sqrt(S) + 1E-10;        

    p = plv .* repmat(pv, nlines, 1);    

    p = p ./ repmat(sum(p, 2), 1, ngroups);  

    pv = sum(p, 1);

    nmembers = zeros(ngroups,1);

    for i = 1:ngroups

        normp = p(:, i) / sum(p(:, i));

        W = diag(normp);

        [eigV, lambda] = eig(A'*W'*W*A);

        [tmp, smallest] = min(diag(lambda));

        v(1:3,i) = eigV(:, smallest);            

        sp = sort(normp, 'descend');

        sp = sum(sp(1:min(length(sp), 2)));

        if (1-sum(sp)) > 0

            sigma(i) = normp' * (l*v(:,i)).^2 / (1-sum(sp)); 

        else

            sigma(i) = Inf;                       

        end

    end        

    % 锟斤拷去锟截革拷锟斤拷

    remove =[];

    for i = 1:ngroups        

        for j = i+1:ngroups

            if (v(1:3, i)'*v(1:3,j) > 0.995) || (pv(i)*nlines <= 3) || (sigma(i) > 10) 

                remove(end+1) = i;

                break;

            end

        end

    end

    if length(remove)>0

        p(:, remove) = [];

        sigma(remove) = [];

        v(:, remove) = [];

        pv(remove) = [];

        ngroups = size(p, 2);

    end

    % 锟斤拷r锟斤拷锟斤拷锟�

    if all(size(v)==size(oldv)) && min(diag(oldv'*v)) > 0.999

        break;

    end    

    oldv = v;

end

v2 = v';

v2(:, 3) = v2(:, 3)+1E-4;

v2 = v2 ./ repmat(v2(:, 3), 1, 3);

v2(:, 1) = v2(:, 1) + imsize(2)/imsize(1)/2;

v2(:, 2) = v2(:, 2) + 1/2;

if nargout > 3

    hpos = vp2horizon(v2, sigma, p, imsize);

end

if DO_DISPLAY

    figure(1), hold off, plot(bincenters, hist_theta, 'b');    

    figure(1), hold on, plot(bincenters, hist_theta, 'r');

    drawnow;

    edge_im = zeros(imsize);

    lines_nnorm(:, [1 2]) = lines(:, [1 2])*imsize(1) + imsize(2)/2;

    lines_nnorm(:, [3 4]) = lines(:, [3 4])*imsize(1) + imsize(1)/2;

    for i = 1:length(groups)

        if length(groups{i})>0

            edge_im = draw_line_image2(edge_im, lines_nnorm(groups{i}, 1:4)', i);

        end

    end

    figure(2), imshow(edge_im); 

    [tmp, bestv] = max(p, [], 2);    

    edge_im = zeros(imsize);

    for i = 1:ngroups

        groups{i} = find(bestv==i);

        if length(groups{i})>0

            edge_im = draw_line_image2(edge_im, lines_nnorm(groups{i}, 1:4)', i);

        end

    end

    figure(3), imshow(255-label2rgb(edge_im));    

    drawnow; 

end

下载 ：