Use of genetic algorithms as detectors of centerlines of local roads on digital images

Abstract

In this work, the use of genetic algorithms is proposed as detectors of straight line segments that characterize a rural road. The segment detection process consists of iterative procedures on digital images, where each step results in two possible segments for the centerlines. The validation of the results produced by the proposed algorithms was based on a manually vectorized reference image. The accuracy index was computed as the percentage of wrongly detected pixels. Despite the modest number of experiments, the results have shown the proposed method to be promising.