Heuristic Bayesian pixel classification for power line inspection

In this paper the heuristic knowledge obtained using the Hough Transform is applied to determine the prior and posterior probabilities for a traditional Bayesian classifier. This novel Bayesian classifier is used in unmanned aerial vehicle (UAV) for overhead power line inspection. The proposed method fuses the Bayesian probabilities and information pertaining to the position and width of the power lines. The results show that this strategy significantly improves the performance of the classifier. This approach presented in this paper can be generalized by using a variety of heuristics, such as contour information of circles, polygons, or other shapes with or without mathematical equations.