Discover MakerZone

MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

Learn more

Discover what MATLAB® can do for your career.

Opportunities for recent engineering grads.

Apply Today

detect and fit curved lines in an image

Asked by Eric Franck on 8 Dec 2012

I'm working on a project that includes lane detection and tracking; I'm able to detect straight lines in image using HT; I'd like to know how to accurately detect and fit curved lines in an image. Thanks folks

0 Comments

Eric Franck

2 Answers

Answer by Image Analyst on 11 Dec 2012

If you have the coordinates of the pixels, you could use polyfit to fit them to a polynomial.

See articles on road following: Vision Bib

3 Comments

Eric Franck on 11 Dec 2012

thanks, the pb is i don't hv those coordinates! i'm able to detect straight lines in binary image using HT without having to manually determine the coordinates of the pixels so I'd like to detect curved lines as well. when the road is flat the algorithm works well but the detection and fitting process become inaccurate when i deal with curved road. it's more like a parabola fitting problem i guess.

Image Analyst on 11 Dec 2012

Well, maybe use the algorithm J recommended above or pick one from the exhaustive list I gave in Vision Bib. Or start simply with an edge detection routine, like the edge() function in MATLAB. Of course getting those coordinates is required before fitting to a curve can be done - you have to have something to fit to.

Image Analyst
Answer by Jürgen on 8 Dec 2012

Hi,

Hough Transform can also be used for detecting curves:http://faculty.washington.edu/cfolson/papers/pdf/cviu99.pdf

regards,J

1 Comment

Eric Franck on 11 Dec 2012

Thanks a lot for your quick reply J, I've read the paper but frankly speaking it's been kinda difficult for me to derive a working code based on that paper within these days. Can u direct me to a working code for curve fitting? not necessarily based on HT of course. the ones I hv can only detect straight lines and no curves!

regards!

Jürgen

Contact us