|M.Sc Thesis||Department of Civil and Environmental Engineering|
|Supervisors:||Assoc. Prof. Shoshany Maxim|
|Prof. Doytsher Yerach|
We live in information flow age. The accessibility, low cost, and the speed in which our world change, increase the demand for an efficient use of available visual sources. Together with the increasing computing capabilities, the environment for building new solutions for automated/semi automated features extraction engines and algorithms is set.
The porous of this thesis is to prove the potential of symmetry in road and elongated feature extraction.
The first step in automated/semi-automated use of symmetry is edge detection.
There for the research focuses on two things:
First we try different kinds of edge detectors to find out that the best results are Canny's algorithm with manual threshold.
In order to achieve the best
results a Gaussian filter is used on the original images. The size and standard
deviation of it are set manually.
After the edge detection is completed, the results are being improved by the Dilation and Erode functions.
Now we skeleton the edges and filter the results by comparing them to the filtered image and other cleaning functions.
Finally the result is compared to manually sampled axes using the dist function.
To some-up the two major contributions of this research are:
Skeletonization is a semi-automated process best used over open areas with low vegetation.