Alphashape boundary detection not working as desired

Question

James Akula 2023-6-22

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1986969-alphashape-boundary-detection-not-working-as-desired

编辑： Matt J 2023-6-27

XY.mat

Given a complex set of (x, y) data I would like to create a hull that is not convex but not fully conforming. I (think I) prefer alphashape over boundary as the former lets me select a specific number of pixels for the radius. My final goal is to create a mask from the outline of the resulting shape. However, my attempt has failed.

Any; suggestions on how to make this work as intended?

load XY.mat

shp = alphaShape(x, y, 100, 'HoleThreshold', 99999999999999);

[~, BXY] = boundaryFacets(shp);

figure

scatter(x, y, '.k')

axis ij equal off

axis([1, max(x) + 500, 1, max(y) + 500])

figure

plot(shp)

axis ij equal off

axis([1, max(x) + 500, 1, max(y) + 500])

% This looks good. I want everything green/black in the final mask.

figure

plot(BXY(:,1), BXY(:,2))

axis ij equal off

axis([1, max(x) + 500, 1, max(y) + 500])

% This seems to have the correct points, but there are lines both

% inside and outside of the desired shape and for it to work right it

% needs to be just the outside boarder of the above

figure

BW = poly2mask(BXY(:,1), BXY(:,2), max(x) + 500, max(y) + 500);

imshow(BW);

% Consequently, the mask looks all wrong.

% I want the mask to to be the entire area enclosed by the alphashape,

% with no holes, and just the area enclosed by the alphashape.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

James Akula 2023-6-23

在 MATLAB Online 中打开

XY.mat

Closer. But still need help...

load XY.mat

shp = alphaShape(x, y, 100, 'HoleThreshold', 99999999999999);

[~, BXY] = boundaryFacets(shp);

subplot(2, 2, 1)

scatter(x, y, '.k')

axis ij equal off

axis([1, max(x) + 500, 1, max(y) + 500])

subplot(2, 2, 2)

plot(shp)

axis ij equal off

axis([1, max(x) + 500, 1, max(y) + 500])

% This looks good

subplot(2, 2, 3)

plot(BXY(:,1), BXY(:,2))

axis ij equal off

axis([1, max(x) + 500, 1, max(y) + 500])

% This looks good, except there are crisscrossing lines and for it to work

% right it needs to be just the outside boarder

subplot(2, 2, 4)

[R, C] = find(ones([max(y) + 500, max(x)+500]));

tf = inShape(shp, C, R);

BW = false([max(y) + 500, max(x) + 500]);

BW(tf) = true;

imshow(BW)

% I was ablt to solve part of my problem--that is, the creation of a

% working mask. But I would still like the border plot (lower left) to

% work, too!

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

James Akula 2023-6-27

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1986969-alphashape-boundary-detection-not-working-as-desired#answer_1263523

在 MATLAB Online 中打开

XY.mat

An answer that seems to work is to convert the alphashape into a triangulation and then use "boundaryshape."

load XY.mat

shp = alphaShape(x, y, 100, 'HoleThreshold', 99999999999999);

subplot(2, 2, 1)

scatter(x, y, '.k')

axis ij equal off

axis([1, max(x) + 500, 1, max(y) + 500])

subplot(2, 2, 2)

plot(shp)

axis ij equal off

axis([1, max(x) + 500, 1, max(y) + 500])

subplot(2, 2, 3)

[tri, P] = alphaTriangulation(shp);

TR = triangulation(tri, P);

polyshp = boundaryshape(TR);

plot(polyshp, 'EdgeColor', 'b', 'FaceColor', 'none')

axis ij equal off

axis([1, max(x) + 500, 1, max(y) + 500])

subplot(2, 2, 4)

ps_regions = regions(polyshp);

BW = false([max(y) + 500, max(x) + 500]);

for ii = 1:numel(ps_regions)

BW = BW | poly2mask(...

ps_regions(ii).Vertices(:,1), ps_regions(ii).Vertices(:,2), ...

max(y) + 500, max(x) + 500);

end

imshow(BW)

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

Answer 2

Image Analyst 2023-6-23

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1986969-alphashape-boundary-detection-not-working-as-desired#answer_1261004

在 MATLAB Online 中打开

Try imfill

BW = imfill(BW, 'holes');

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

James Akula 2023-6-23

Thank you for this advice, but filling the holes isn't the solution, as that would do "too much." The "outline" of the blue plot does not match that of the alphashape, as lines crisscross both internally and externally.

请先登录，再进行评论。