How do I join the points of a scatter plot to make a line?

I have a scatter plot as given below. Each point comes from a xy pair (for example -

         670         358
         686         363
        1677         365
         681         369
        1680         376
         724         377

How do i join these points to make a line?

Answer 1

[Matt J]

Use PLOT instead of SCATTER. Connecting with lines defeats the purpose of a scatter plot.

[Coder Bee]

Plot and scatter give me the similar image. Not connecting these points defeats the purpose of my goal. I need to connect them to form a line/

[Matt J]

You can tell PLOT to connect the points with lines, e.g.,

plot(1:5,'--*');

[Steven Lord]

Are you using just a marker specification or a marker specification plus a color specification in your plot call? If so, you need to add in a line style specification.

[Coder Bee]

When I use plot or line - it becomes like this. I need to join the xy pairs as in the scatter plot.

[Matt J]

Using your example data

xy =

         670         358
         686         363
        1677         365
         681         369
        1680         376
         724         377

I would do

xys=sortrows(xy,1);  
plot(xys(:,1), xy(:,2),'--o'); 
  ylim([0,600])

[Coder Bee]

I need the line to follow the same shape as the points. I am basically trying to extract the points of a hand-drawn line and recreate it on matlab using the features.

[Matt J]

I think you mean you are trying to find a best fit line to the points.

[Coder Bee]

Yes, can you help me?

Answer 2

[Walter Roberson]

If you have R2014b or later, use boundary to get the outline of the implied shape. If you are using an earlier release you can look in the File Exchange for Alpha Shapes routines.

"following the same shape as the points" is not well defined when there are multiple points with the same X or same Y coordinate.

[Coder Bee]

I use R2014a. Alpha shapes doesn't seem to be working well for me. Can you help me with an approach of replicating a hadn-drawn line on matlab?

[Walter Roberson]

http://www.mathworks.com/matlabcentral/fileexchange/28851-alpha-shapes

http://www.mathworks.com/matlabcentral/fileexchange/6760-ashape--a-pedestrian-alpha-shape-extractor

You have not presented a hand-drawn line that we could examine the properties of. We need to observe the decision logic of how to connect points that are close together.

Answer 3

[Image Analyst]

See attached demo for polyfit. It will show you how to fit a line through your points.

[Matt J]

polyfit is not the best choice because both x and y have measurement noise in them.

[Image Analyst]

I bet it would do a respectable job though. Maybe not the theoretical best, but usable for his purposes. Perhaps if he uploaded the data we could try it.

Answer 4

[Matt J]

Here's a total least squares fit,

%Get line equation

xymean=mean(xy);
[~,~,V]=svd( bsxfun(@minus,xy,xymean) ,0);

e=V(:,end);
 e=[e;-dot(e,xymean)];

fun2=@(x,y) e(1)*x+e(2)*y +e(3);

%Plot

scatter(xy(:,1), xy(:,2) ); hold on

xymin=min(xy);
xymax=max(xy);
xb=[xymin(1), xymax(1)];
yb=[xymin(2), xymax(2)];

ezplot(fun2,[xb,yb]); hold off

[Coder Bee]

I need the data to join into a line which looks like this.

[Walter Roberson]

In the center part of this sample, there is an area in which there is a loop. I cannot see any justification for a loop being there based upon the original image. I also cannot justify the large gap in the line being where you show it -- not unless there are additional gaps at least as large.

[Matt J]

@Coder,

For the sake of clear communication, you need to distinguish between lines and curves. The thing you've drawn is a very wavy curve, not a line. A line, by definition, is perfectly straight.

Nor does it seem to be a "fit" as you said you wanted earlier. It looks like the exact drawing that you started with. In that case, why not just use the original image that you are given for whatever it is that you're doing?

If you're trying to reconstruct the exact path followed by the pen, you won't be able to without further information and constraints. There is no unique path that connects a given discrete set of points.