Horizon line detection
This is an implementation of a relatively basic horizon line detection based on dynamip programming and learning. There are 3 different weights for each node (pixel), low (point detected as potential horizon line), medium (edge), and high (any other). Ideally the entire horizon line will be detected by the features and the path will be found quickly. In many cases the edges in the horizon lines are not correctly detected as horizon line so an edge can be take. In few caes the horizon line is not even seen as an edge so we need to explore the area but there's a maximum explorarion steps that could be taken.
The algorithm is very similar to:"A Machine Learning Approach to Horizon Line Detection Using Local Features", T Ahmad, G Bebis, EE Regentova, A Nefian - Advances in Visual Computing, 2013
Usage for train: ./HorizonLineDetector 0 file_path_in training_file
Usage for run : ./HorizonLineDetector 1 file_path_in file_path_out
file_path_in will be a list file in the case of training and a video file path in the case of running
file_path_out video file path
training_file will config file output in the case of training and will determine the file to use when running
