Detection of Non-structured Roads Using Visible and Infrared Images and an Ant Colony Optimization Algorithm

In this paper, the problem of road detection on unstructured environments is mapped as an optimization one, and an Ant Colony Optimmization algorithm is applied to solve it. This allows that the artificial agents make up for the lack of edge input information with their common memory. This common memory contains the paths followed by the entire colony. From an intuitive point of view, it can be said that in the areas where no edge information can be obtained, the agents will construct pheromone bridges (to the next edge detected pixels), that will allow them to continue constructing their solution, achieving this way a robust road detection. The input information for the algorithm will consist in images captured by cameras in the visible and in the infrared spectrums. Some features of the roads that are wanted to be detected make that in some cases the information from the visible spectrum is more useful than the information from the infrared spectrum while in other cases the infrared ones are more useful. This is why both sources of data are used to improve the global performance of the proposed method.