Context: I want to create an interactive heatmap with areas separated by a ZIP code. I've found no way of displaying it directly (i.e. using Google Maps or OSM), so I want to create curves or lines that are separating those areas, and visualize it in maps.
I have a set of points, represented by their coordinates and their according class (ZIP code). I want to get a curve separating them. The problem is that these points are not linearly separable.
I tried to use softmax regression, but that doesn't work well with non-linearly separable classes. The only methods I know which are able to separate non-linearly are nearest neighbors and neural networks. But such classifiers only classify, they don't tell me the borders between classes.
Is there a way to get the borders somehow?
If you have a dense cloud of known points within each Zip code with coordinates [latitude. longitude, zip code], using machine learning to find the boundary enclosing those points sounds like overkill.
You could probably get a good approximation of the boundary by using computational geometry, e.g finding the 2D convex hull of each Zip code's set of points using the Matlab convhull function
K = convhull(X,Y)
The result K would be a vector of points enclosing the input X, Y vector of points, that could be used to draw a polygon.
The only complication would be what coordinate system to work in, you might need to do a bit of work going between (lat, lon) and map (x,y) coordinates. If you do not have the Matlab Mapping Toolbox, you could look at the third party library M_Map M_Map home page, which offers some of the same functionality.
Edit: If the cloud of points for Zip codes has a bounding region that is non convex, you may need a more general computational geometry technique to find a better approximation to the bounding region. Performing a Voronoi tesselation of the region, as suggested in the comments, is one such possibility.
Related
In part of an Artificial Neural Network matlab code, I want to find nearest points of two convex polygons.
I saw
dsearchn(X,T,XI)
command's description here, but that finds closest points between two sets of points, and polygons (like convexs) have infinites points.
So can you suggest any way/idea?
Notice: I'm using MATLAB 2014a. I have the coordinate of each convex's vertex point.
If you are not happy with what is provided by dsearchn, then, If I were you, I would do one of two following:
Find Nearest Neighbours on the vertices (for example which vertex of
polygon A is the NN of a given vertex of polygon B).
Pick a random point inside polygon A (you may want to compute the
convex hull of A, but you may skip that and take into account only
the vertices you know already). That random point is the query. Find
an NN of that point from the vertices of polygon B.
You may want to ask in Software recommendations for more.
Edit:
Another approach is this:
Create a representative dataset of polygon A. Set the size of the dataset yourself and fill it with samples of points that lie inside the polygon. Choose them uniformly randomly inside the polygon.
Then take a point of polygon B (either a vertex or a random point inside polygon B) and that's the query point, for which you will seek Nearest Neighbour(s) inside the representative dataset of polygon A.
Of course that's just an approximation, but I can't think of something else now.
Notice that you can of course do the same for polygon B.
With This File in File Exchange, I've find the solution.
With a little code modification, I have drawn a perpendicular bisector which I wanted. Of course, this way is time consuming.
I need some help with matlab and detected features. I'll start with the fact that I am not so good with matlab.
I am trying to automate morphing process by combining feature point detection with morph. I use CascadeObjectDetector to find the features of a human face for two images and I find their strongest features and position with Location function. What I need is to sort the x,y coordinates of detected features so that coordinates of points in pic1 would match the nearest coordinates of pic2 because the main thing in morphing is to have pairs of corresponding points.
I am working with structured 2.5D and unstructured 3D data, which generally is available in (X,Y,Z) coordinates, i.e. point clouds. Now I want to impose a regular voxel grid onto the data. This is not meant for visualization purposes, but rather for "cleaning" or fusing the data. I imagine cases, where e.g. 3 points fall within the volume of one voxel. Then I would aim at either just setting this voxel to "activated" and discarding the 3 original points or alternatively I would like to calculate the euclidian mean of the points and return the thus "cleaned" point cloud as an irregularly structured one again.
I hope I could make my intentions clear enough: It's not about visualization and not necessarily about using volumetric cubes instead of points. It's only about manipulating possibily irregular point clouds in a structured way.
I was thinking about kd-tree or octree based solutions in this context, but can anybody point me in the proper direction? Hinting at existing MATLAB implementations would be most appreciated.
If the data is irregularly spaced, what you want to use is something which both smooths and interpolates your data points. A very good method for doing so is Gaussian process regression. Here's an example for the same problem but in 2D.
I have 3D space. And I know for example N points in this space (x1,y1,z1), (x2,y2,z2)..., (xn,yn,zn). I want to interplolate points, that is different from this. How can I do this in Matlab?
interp3 may help you. Here is the documentation.
As always, there are questions left unanswered by your one line query.
If the data is of the form where there is a functional relationship z(x,y), (or y(x,z) or x(y,z)) then you might potentially be able to use one of the interpolation tools. Thus, suppose you have data that lies on a lattice in the (x,y) plane, thus some value of z at each point in that lattice. In this case, you can use interp2.
Alternatively, if the data is scattered, but there is some single valued functional relationship z(x,y) that you don't have, but it is some continuous function. Infinite first derivatives are a problem too here. In this case, assuming that you have data that at least fills some convex domain in the (x,y) plane, you can still interpolate a value of z. For this, use griddata, or TriScatteredInterp. Or you might use my own gridfit tool, as found on the file exchange.
Next, the way you describe the data, I'm not at all positive that you have something in one of the above forms. For example, if your data lies along some curved path in this 3D domain, and you wish to interpolate points along that curved arc can be done using my interparc tool, also found on the file exchange.
One last case that people often seem to have when they talk about interpolation of a spatial set like this, is just a general surface, that they wish to build a neatly interpolated, smooth surface. It might be something as simple as the surface of a sphere, or something wildly more complex. (These things are never simple.) For this, you might be able to use a convex hull to approximate something, if it is a closed convex surface. More complex surfaces might require a tool like CRUST, although I have no implementation of it I can offer to you. Google will help you there, if that is what you need.
The point of all this is, how you interpolate your data depends on what form the data is in, what it represents, and the shape of the relationship you will be interpolating.
the output of some processing consists of a binary map with several connected areas.
The objective is, for each area, to compute and draw on the image a line crossing the area on its longest axis, but not extending further. It is very important that the line lies just inside the area, therefore ellipse fitting is not very good.
Any hint on how to do achieve this result in an efficient way?
If you have the image processing toolbox you can use regionprops which will give you several standard measures of any binary connected region. This includes
You can also get the tightest rectangular bounding box, centroid, perimeter, orientation. These will all help you in ellipse fitting.
Depending on how you would like to draw your lines, the regionprops function also returns the length for major and minor axes in 2-D connected regions and does it on a per-connected-region basis, giving you a vector of axis lengths. If you specify 4 neighbor connected you are fairly sure that the length will be exclusively within the connected region. But this is not guaranteed since `regionprops' calculates major axis length of an ellipse that has the same normalized second central moment as the connected region.
My first inclination would be to treat the pixels as 2D points and use principal components analysis. PCA will give you the major axis of each region (princomp if you have the stat toolbox).
Regarding making line segments and not lines, not knowing anything about the shape of these regions, an efficient method doesn't occur to me. Assuming the region could have any arbitrary shape, you could just trace along each line until you reach the edge of the region. Then repeat in the other direction.
I assumed you already have the binary image divided into regions. If this isn't true you could use bwlabel (if the regions aren't touching) or k-means (if they are) first.