multiple line fitting on scatter plot

Hi,

I have a large amount of data from which i made scatter plots. The data has distinct features through which the scatter could be devided into 3 parts. Stable in the beginning from 0 to 0.3, then constant decline from 0.3 to 1 and then constant increase from 1 to 3.

Now i want to get 3 straight lines from these 3 areas, and also the lines should be connected. I tried curve fitting but it does not solves the problem as the break points are important indicators. I tried some tools and scripts shared on this forum, but all of them do not allow control over the placement of knots (or may be they allow control but i could not find out how to do that).

Answer 1

[Image Analyst]

You might check out the minimum perimeter polygon algorithm. http://dip.sun.ac.za/~hanno/tw444/lesings/lesing_19.pdf

Answer 2

[KSSV]

doc lsline.

[Bilal Babar]

The issue with using lsline for multiple portion is that the line extend beyong the boundaries of the the sections. and i could not find a way to make the separate lines connect with each other.

Answer 3

[Image Analyst]

Try RANSAC https://en.wikipedia.org/wiki/Random_sample_consensus