Object Description Based on Local Features Repeatability

This chapter describes a methodology for selecting characteristic local features, to facilitate the detection of specified objects on power supply lines in diversified scenes. The objective of this works was to automate a process intended to search for characteristic objects in digital images, utilizing a limited training set. Description of the insulator in the test was based on a set of distinctive local features, which in the individual images of the test set, demonstrated the highest degree of repeatability. Such set allowed the location of insulators in other samples obtained from a diagnostic flight. To verify the method, the verification sets were proposed. To search and describe local features the SURF algorithm was used. The elaborated training set and adopted methodology of the test provided promising results within the analyzed scenario (flight along the power supply line). Analyses that were carried out, aimed at the development of methods supporting the power network maintenance, basing on the advanced techniques of digital image processing.