On a method for supporting visual identification of underwater objects

Abstract

Abstract The search and detection of objects under water is carried out by groups of specialised divers. However, their time underwater and their ability to penetrate the depths are limited. For these reasons, the use of unmanned underwater vehicles equipped with technical observation equipment, including TV cameras, is becoming increasingly popular for these tasks. Video images from cameras installed on vehicles are used to identify and classify underwater objects. The process of recognition and identification of objects is tedious and difficult and requires the analysis of numerous sequences of images, and so it is desirable to automate this process. In response to these needs, this article presents the concept of identification of underwater objects based on visual images from an underwater body of water sent from an unmanned underwater vehicle to a base vessel. The methods of initial processing of the observed images from an underwater area as well as the method of searching for selected objects in these images and their identification with the use of the Hough transform will be described. Furthermore, the paper presents the results of the preliminary processing and identification of the observed images following a deconvolution operation.