How to reduce data irregularly?

Hi All,

I have a plot consists of 10000 points that I need to reduce to upload it on an instrument which accepts only 200 points. The plot consists of three parts: first a straight line, a curve and again a straight line. I want to reduce the number of points like getting a small number of points for both lines and get more points for the curve. How can I do that?

[Image Analyst]

Can you attach your data in a .mat file with the paperclip icon?

Answer 1

[Star Strider]

One approach would be to use the gradient function to calculate the derivative of the available data. This will be greater than 1 (or different from a constant, depending on what the striaght lines are). Then randomly select points from the identified curve, and most likely adjacent points on the line (for example the last 5 before the curve begins and the first 5 after itt is defined to end). (These would be sorted by their independent variable values so the independent variable values are increasing.)

The gradient function is preferable to diff because the output is the same size as the input, making indexing easier.

.

[Mohamed Ibrahim]

Thanks for your reply but I have some questions. To clarify more the situation, my plot is shown here

so the imported data is stored in two vectors VarName 1 and VarName2. I should use gradient with the following command: gradient(VarName1, VarName2) right?

If it is so, I got the following plot:

Then I will start to choose 5 points before the curve and after the curve? how will they still be reduced?

[Star Strider]

This is not the sort of plot I expected to see.

I have no idea how to deal with that, other than to use:

idx = fix(linspace(1, 1E+4, 200));  

idx = 1×200

1 51 101 151 201 252 302 352 402 453 503 553 603 654 704 754 804 855 905 955 1005 1056 1106 1156 1206 1257 1307 1357 1407 1458

and use ‘idx’ as the indices into both vectors. Those data (not supplied so I have no idea what they actually are) appparently have no specific statistical characteristics in the approximations I made to them that could be used for any sort of selective sampling.

.

Answer 2

[Kelly Kearney]

It sounds like you're looking for a line simplification algorithm. These algorithms try to preserve the shape of a polygon or polyline while reducing the number of points used. The Mapping Toolbox includes a version of the Douglas-Peucker method (reducem), and it looks like there are some FEX options (maybe this one?) available as well if you don't have that toolbox.

Answer 3

[Jan]

See: https://www.mathworks.com/matlabcentral/answers/1454924-downsample-data-adapively-intelligently#answer_789429