Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

MATLAB - best way to dynamically update a line handles' XData and YData?

Tags:

matlab

figures

I am collecting data and plotting that data in real time. The data are produced by a motion capture system. I have one class DynamicDataset that is just a wrapper around a 2-column matrix (although it's more nuanced than that) with an event notifier for new data added; another class DynamicPlotter that listens for the data-added event and updates the plot dynamically. Appropriate code snippets:

classdef DynamicDataset < handle
    properties
        newestData = [];
        data = []
    end
    events
        DataAdded
    end
    methods
        function append(obj, val)
            obj.data(end+1,:) = val;
            obj.newestData = val;
            notify(obj, 'DataAdded');
        end
    end
end

classdef DynamicPlotter < dynamicprops
    properties
        FH %# figure handle
        AH %# axes handle
        LH %# array of line handles - may have multiple lines on the plot

        dynProps = {} %# cell array of dynamic property names - 
                      %# use to access individual datasets
    end
    methods
        function obj = DynamicPlotter(props) %# props is a cell array of dynamic 
                                             %# properties to store information
            for i = 1:length(props) 
                addprop(obj, props{i});
                obj.(props{i}) = DynamicDataset;
                obj.dynProps = [obj.dynProps props{i}];

                addlistener(obj.(props{i}), 'DataAdded', @obj.updatePlot(i));
            end
            obj.createBlankPlot();
        end

        function createBlankPlot(obj)
            obj.FH = figure;
            obj.AH = axes;

            hold all;

            for i = 1:length(obj.dynProps)
                obj.LH(i) = plot(nan); %# only used to produce a line handle
                    set(obj.LH(i), 'XData', [], 'YData', []);
            end
        end

        function updatePlot(obj, propNum)
            X = get(obj.LH(propNum), 'XData');
            Y = get(obj.LH(propNum), 'YData');

            X(end+1) = obj.(dynProps{propNum}).newestData(1);
            Y(end+1) = obj.(dynProps{propNum}).newestData(2);

            set(obj.LH(propNum), 'XData', X, 'YData', Y);
        end
    end
end

Based on the MATLAB Code Profile, the set command in updatePlot() is rather expensive. I am wondering if there is a better way to plot individual points as they come? Ideally I would push the single point into XData and YData and draw that point only, but I don't know if this is possible.

Please note that there may be multiple lineseries objects (i.e., multiple graphs on the same plot); plot() takes an axes handle as an argument, so it wouldn't consider the properties of the previously drawn line handles (or is there a way to make it do so?); I thought of just doing plot(x,y);hold all; but that would give me separate line handles every time, each corresponding to a single point.

It might be that there's no way to make plotting incoming points any faster, but I figured I'd ask.

EDIT: Updated OP with actual code I'm working with, rather than using a generic example that's up for misinterpretation.

like image 212
Dang Khoa Avatar asked Sep 14 '11 08:09

Dang Khoa


2 Answers

The amount of data you're handling in each update, is large (although only a single point is actually changing), making your code O(N^2).

By using a second lineseries to build up a large group of data, you can alternate between adding every point to a short "active" line, and infrequently adding large blocks to the main lineseries. While this doesn't exactly avoid O(N^2), it lets you reduce the constant significantly.

If you do this, remember to overlap the "old" lineseries and "active" lineseries by one point, so that they connect.

Essentially:

    function updatePlot(obj, propNum)
        X = get(obj.LHactive(propNum), 'XData');
        Y = get(obj.LHactive(propNum), 'YData');

        X(end+1) = obj.(dynProps{propNum}).newestData(1);
        Y(end+1) = obj.(dynProps{propNum}).newestData(2);

        if numel(X) > 100
            Xold = [get(obj.LH(propNum), 'XData'); X(2:end)];
            Yold = [get(obj.LH(propNum), 'YData'); Y(2:end)];
            set(obj.LH(propNum), 'XData', Xold, 'YData', Yold);

            X = X(end);
            Y = Y(end);
        end

        set(obj.LHactive(propNum), 'XData', X, 'YData', Y);
    end
like image 163
Ben Voigt Avatar answered Oct 31 '22 08:10

Ben Voigt


Part of the reason why your code may be taking a long time to run is because you are using a for loop to assign your variables. Depending on what version of Matlab you are using, this will slow your process down significantly. I suggest using vectorization to assign values to your x and y like this:

x = 1:1000;
y = cosd(x);

You can then assign the first points in your data.

xi = x(1);
yi = y(1);

When you plot, assign the XDataSource and YDataSource.

h = plot(xi, yi, 'YDataSource', 'yi', 'XDataSource', 'xi');

Now when you loop through to change the values, use the refreshdata to update the Xdata and Ydata values. Use the drawnow function to update the figure window.

for k = 2:1000,
xi = x(1:k);
yi = y(1:k);
refreshdata(h, 'caller')
drawnow;
end
like image 42
Thea Avatar answered Oct 31 '22 07:10

Thea