I have a 3D space which I discretized in voxels (cubes of volume). I also have a set of 3D points in such space. I want to know the expected number of points in a given voxel. I chose GMM as a model for this purpose, but I do not know how to calculate what I want starting from mu, sigma and weight of each Gaussian.
So far I managed to fit the GMM (easy):
obj = gmdistribution.fit(points', 20);
and I to plot it via
figure(1);
hold on;
for i = 1:k
plot_gaussian_ellipsoid(obj.mu(i,:), obj.Sigma(:,:,i));
end
axis equal;
which results in what I expect, that is a map where the colors tell me the concentration of points.
The question is, how can I extract the expected number of points in a voxel, given its center (x,y,z) and its side s?
You can use (see example here http://www.mathworks.nl/help/stats/gmdistribution.cluster.html)
idx = cluster(gm,points);
Related
I have this data with values on the edges of the matrix and other values at evenly spaced interval within the matrix. I want to predict the values of the zero positions from the original values and make a heat map of the new data. Through suggest, I use scatteredInterpolant, ndgrid and interpolant since the data is that interp2 (matlab functions) cannot be used to interpolate the zero elements. Now, this method doe not give me a smooth figure and I am want to know if someone can offer some help. I have attached the figure from my code, the data and the code to this post.Thank you.
[knownrows, knowncolumns, knownvalues] = find(DataGrid); %get location and value of all non-zero points
interpolant = scatteredInterpolant(knownrows, knowncolumns, knownvalues,'linear'); %create interpolant from known values
[queryrows, querycolumns] = ndgrid(1:1:size(DataGrid, 1), 1:1:size(DataGrid, 2)); %create grid of query points
interpolatedj = interpolant(queryrows, querycolumns);
HeatMap(interpolatedj)
https://www.mediafire.com/?pq40x1ljxk8h996
https://www.mediafire.com/?pq40x1ljxk8h996
To plot a smoothed matrix you can use pcolor and set the shading parameter to interp
pcolor(M); %where M is your 2D matrix
shading interp %set the shading to interp
Try
image(M) or imagesc(M) where M is a matrix. pcolor(M) also works. If your matrix is huge then you need to remove edges otherwise figure just looks like blank image.
What I want to do is very simple, I just cannot seem to get MATLAB to do it. I would like to plot contours using my 2D data set.
My data set is large; 2 x 844240. I can do a scatter plot just fine,
scatter(Data(1,:), Data(2,:));
Reading through the forums I found Scatter plot with density in Matlab, where a hisogram was plotted. This would suffice, however, I would like to overlay the plots.
The issue is that they have different axis, my scatter data has an axis of [0 0.01 0 2500]; whereas the histogram is [0 100 0 100].
Is there a way to change the axis values of the histogram without modifying the image?
Thanks!
If I understand correctly, you are using hist3 to construct a histogram and then using imagesc to plot it. You can use the second output argument of hist3 to get the histogram bin centers, and then pass those on to imagesc, e.g.
nBins_x = 100;
nBins_y = 100;
[counts, bin_centers] = hist3(Data, [nBins_x nBins_y]);
x_bin_centers = bin_centers{1};
y_bin_centers = bin_centers{2};
imagesc(x_bin_centers, y_bin_centers, counts)
A couple other notes:
In your case, you will need to transpose your [2 x N] matrix when passing it to hist3, which expects an [N x 2] matrix.
imagesc puts the first axis (which I've been calling the "x" axis) on the vertical axis and the second on the horizontal axis. If you want to flip it, you can use:
imagesc(y_bin_centers, x_bin_centers, counts')
If you want to specify the histogram bins explicitly (e.g. to match your scatterplot) you can specify that in the arguments to hist3:
x_bin_centers = linspace(0, .01, 100);
y_bin_centers = linspace(0, 2500, 100);
counts = hist3(Data, {x_bin_centers, y_bin_centers};
And if you want a contour plot, you can use (note that contour takes the axes arguments in a different order than imagesc):
contour(x_bin_centers, y_bin_centers, counts');
If you are unhappy with the jaggedness of the contours, you may consider using a kernel density estimate instead of a histogram (check out ksdensity) (oops, looks like ksdensity is 1-D only. But there are File Exchange submissions for bivariate kernel density estimation).
I am looking for help for my particular problem.
I have a contour plot created from XYZ data. This plot contains 2 broad peaks with one more intense than the other.
When the most intense peak is aligned with the Y axis, I can perform a fitting of every YZ curve at each X values. I usually do a gaussian fit to plot the peak center on the same graph.
In some cases I need to perform the same fitting but no along the Y axis direction (in this case I just plot YZ scan at every different X values) but along another arbitrary direction.
For the moment the only way I found is the following:
-plot the contour plot and find for the position of the most intense peak
-if the position is not aligned with the Y axis, then rotate all the datas and plot again the contour
-perform the YZ gaussian fit for every X value
- Rotate the resulting XY position to go back to the original plot
-plot the XY position as a line on the original contour plot
this is quite long and requires a lot of memory. i would like ot know if there is a more elegant/faster way.
Thanks for your help
David
I take it you want to extract data from the (x,y,z) data along some arbitrary line in order to make a fit. A contour plot will show only part of the data, the full z(x,y) data can be shown with imagesc etc. Say you want the data along line defined by two points (x1,y1) -> (x2,y2). According to the eq of the line, the line y=a*x+b the slope a is (y2-y1)/(x2-x1) and b=y1-a*x1. For example, I'll select (x,y) coordinates in the following contour:
Create data and end points:
m=peaks(100);
x1=11 ; x2=97;
y1=66; y2=40;
Thus the line parameters are:
a=(y2-y1)/(x2-x1);
b=y1-a*x1;
and the line is:
x=x1:x2;
y=round(a*x+b);
select the proper (x,y) elements using linear indexing:
ind=sub2ind(size(m),y,x)
plot:
subplot(2,1,1)
contour(m,10); hold on
line([x1 x2],[y1 y2],'Color',[1 0 0]);
subplot(2,1,2)
plot(m(ind))
You can now use vec=m(ind) to fit your function.
I have a 2d image, I have locations where local minimas occurs.
I want to measure the width of the valleys "leading" to those minimas.
I need either the radii of the circles or ellipses fitted to these valley.
An example attached here, dark red lines on the peaks contours is what I wish to find.
Thanks.
I am partially extending the answer of #Lucas.
Given a threshold t I would consider the points P_m that are below t and closer to a certain point m of minimum of your f (given a characteristic scale length r).
(You said your data are noisy; to distinguish minima and talk about wells, you need to estimate such r. In your example it can be for instance r=4, i.e. half the distance between the minima).
Then you have to consider a metric for each well region P_m, say for example
metric(P_m) = .5 * mean{ maximum vertical diameter of P_m ,
maximum horizontal diameter of P_m}.
In your picture metric(P_m) = 2 for both wells.
On the whole, in terms of pseudo-code you may consider
M := set of local minima of f
for_each(minimum m in M){
P_m += {p : d(p,m) < r and f(r)<t} % say that += is the push operation in a Stack
}
radius_of_region_around(m) = metric(P_m); %
I would suggest making a list of points that describe the values at the edge of your ellipse, perhaps by finding all the points where it crosses a threshold.
above = data > threshold
apply a simple edge detector
edges = EdgeDetector(above)
find coordinates of edges
[row,col] = find(edges)
Then apply this ellipse fitter http://www.mathworks.com/matlabcentral/fileexchange/3215-fitellipse
I'm assuming here you have access to the x, y and z data and are not processing a given JPG (or so) image. Then, you can use the function contourc to your advantage:
% plot some example function
figure(1), clf, hold on
[x,y,z] = peaks;
surf(x,y,z+10,'edgecolor', 'none')
grid on, view(44,24)
% generate contour matrix. The last entry is a 2-element vector, the last
% element of which is to ensure the right algorithm gets called (so leave
% it untouched), and the first element is your threshold.
C = contourc(x(1,:), y(:,1), z, [-4 max(z(:))+1]);
% plot the selected points
plot(C(1,2:end), C(2,2:end), 'r.')
Then use this superfast ellipse fitting tool to fit an ellipse through those points and find all the parameters of the ellipse you desire.
I suggest you read help contourc and doc contourc to find out why the above works, and what else you can use it for.
I have a 2D scatter plot in MATLAB. Is it possible to interpolate the scatter plot to create an area plot?
If you're simply trying to draw one large filled polygon around your entire set of scattered points, you can use the function CONVHULL to find the convex hull containing your points and the function PATCH to display the convex hull:
x = rand(1,20); %# 20 random x values
y = rand(1,20); %# 20 random y values
hullPoints = convhull(x,y); %# Find the points defining the convex hull
patch(x(hullPoints),y(hullPoints),'r'); %# Plot the convex hull in red
hold on; %# Add to the existing plot
scatter(x,y); %# Plot your scattered points (for comparison)
And here's the resulting figure:
Scatter is generally used to represent data where you can't use a line graph, i.e., where each x might have many different y values, so you can't convert directly to an area graph--it would be meaningless. If your data actually is representable as a line graph, then pass it to area directly.
So I'm not quite sure what you want, but here are some possibilities:
You could create a Voronoi diagram based on your points. This will show a region near your points showing which points are closer to a specific point: voronoi(x,y), or see the help.
You could bucket or quantize your data somehow, making it fit into a grid, and then plot the grid. This could also be considered a histogram, so read up on that.
You could just use larger scatter markers (scatter(x,y,scale) where scale is the same dimensions as x and y).