CODEWITHSUNDEEP
matlabmatrixlarge-data
Amitava
Amitava

Reputation: 431

Combining overlapping matrix data to create a single matrix

I have measurement data from four sensors, each gives current speed at a given water depth (d) and time (t). Below is the matrix dimension of these four current measurements:

cs1 = [d1 x t1]; cs2 = [d2 x t2]; cs3 = [d3 x t3]; cs4 = [d4 x t4]

The water depth arraysd1, d2, d3, d4 are unique values, but their range overlaps. For example, d1=5:4:141 and d2=72:2:200. Time arrays t1,t2,t3,t4 also have different start and end points with different timesteps.

I need to create a combined matrix cs which includes data from all four sensors. For this I have created a NaN matrix as:

t = unique([t1;t2;t3;t4]);
d = unique([d1 d2 d3 d4]);
cs = NaN(length(d),length(t));

% Populaitng data from 1st sensor
for i=1:length(d1)
    for j=1:length(t1)
        cs(d==d1(i) & t==t1(j))=cs1(i,j);
    end
end

I am using a for loop to populate the cs matrix for one instrument data at a time. This way is quite inefficient since both depth and time array sizes are quite large.

Is there a quicker method to fill the cs matrix without running a for loop for each intrument?

Upvotes: 0

Views: 39

Answers (1)

verbatross
verbatross

Reputation: 607

First we generate some sample input data. Note that this data is in a data structure that allows us to loop over the sensors, instead of writing separate code for each of the four sensors.

% sample input data
sensors = struct;
sensors.d = { ...
    1:1:5; % d1
    2:2:6; % d2
    3:2:7; % d3
    2:1:4; % d4
    };
sensors.t = { ...
    1:2:7; % t1
    2:3:5; % t2
    1:4:9; % t3
    3:1:5; % t4
    };
sensors.cs = { ...
    1 * ones(length(sensors.d{1}), length(sensors.t{1})); % cs1
    2 * ones(length(sensors.d{2}), length(sensors.t{2})); % cs2
    3 * ones(length(sensors.d{3}), length(sensors.t{3})); % cs1
    4 * ones(length(sensors.d{4}), length(sensors.t{4})); % cs2
    };

Then we combine the sensor data into a single matrix. It's not clear from your example how you want to combine the data, so we will make the assumption that you want the combined matrix to start as all zeros and then add each sensor's data element-by-element. 

% COMBINE SENSOR DATA INTO SINGLE MATRIX

% dimensions
d = unique(cat(2, sensors.d{:}));
t = unique(cat(2, sensors.t{:}));

% initialize matrix
cs = zeros(length(d), length(t));

% loop over sensors
for s = 1 : length(sensors.cs)

    % indexes of sensor's d and t dimensions in combined d and t dimensions
    [~, Id] = ismember(sensors.d{s}, d);
    [~, It] = ismember(sensors.t{s}, t);

    % add the values to the combined matrix
    cs(Id, It) = cs(Id, It) + sensors.cs{s};

end

If instead you want to, for example, start from an all-NaN matrix and have each element represent the last sensor whose data is added, you can replace the line cs = zeros(length(d), length(t)); with cs = NaN(length(d), length(t)); and the line cs(Id, It) = cs(Id, It) + sensors.cs{s}; with cs(Id, It) = sensors.cs{s};.

Upvotes: 1

Related Questions