Road detection in satellite images

Hi, As a part of my project, I have implement a road detection mechanism that takes in grayscale satellite images and produces an output that detects all the straight, curved and narrow roads. I have implemented a crude version of this using hough transform but I have not been able to achieve satisfactory results. If any of you could help me with a better algorithm I would really appreciate it and it would allow me to proceed to the next stage of my project.

Thank you.

[rizwan]

Hi Yash,

I am working on the same Road detection, can you please share your mat lab code for road detection. I shall b thank full as i can move forward from your help

best regards

[PRATIK MISHRA]

Hello, First of all plot the histogram equivalent of the input satellite image. Then select that region in your histogram plot where the approximated road lies. Then apply a mask of that range to your image after converting it to black and white. Then do morphological opening and closing and then morphological thinning.

[Darsana P M]

Is the matlab code available for road detection?

[Image Analyst]

The primitive and simple methods described above will not be very robust - they cannot be used on a wide variety of images. They're just too basic. I suggest you look at the papers that were published on the topic like in the link I gave below, and contact the authors to see if you can buy the MATLAB code from them, if they have any in that language.

Answer 1

[Image Analyst]

We help with MATLAB questions on syntax, error messages, and simple program flow, not complicated algorithm development. You can find algorithms here in Vision Bib: http://www.visionbib.com/bibliography/contentscartog.html#Cartography,%20Aerial%20Images,%20Remote%20Sensing,%20Buildings,%20Roads,%20Terrain,%20ATR