Path tracking in underwater vehicle navigation - on-line evaluation of guidance path errors via vision sensor

In this paper a vision-based algorithm for on-line estimation of position control errors in a guidance system for path tracking is presented. The algorithm uses techniques of pattern recognition, however with a high degree of morphological simplifications. The algorithm detects the path line stretch which is displayed in the actual frame, and estimates the slope of the line and its midpoint. Moreover, it identifies a confidence zone in where the line stretch would be certainly. Ad-hoc experiments with a subaquatic vehicle in an test tank show the features of the algorithm proposed under strong conditions of light perturbation and cloudy water.