Detecting faults using video streams from drone inspections

DNV GL - Maritime Classification in Poland would like to use drones in ship hull survey. This has the potential for accelerating the survey and reducing staging is becoming visible

“The idea to test drones to support surveying a ship tanks was motivated by our aim to accelerate the survey work and by the more widely available drone technology” says Cezary Galinski, the flying squad manager based in Gdansk.

Visual inspection is the only effective way of assessing the structural condition of large merchant vessels. Currently, the surveyor has to perform clos-up visual inspection to assess the condition of coating and to detect cracks which have to be repaired. Scaffolding has traditionally been the only means to get access to the critical areas of large tanks which can be as large as 30x30x30 meters.

The basic approach is to use the camera-equipped drone to visually evaluate remote structural components for checking their condition. By using the video stream, the surveyor can remotely review the condition of a structure.

We would like to take this approach a step further where anomalies can be detected automatically through analysis of the video stream. This will allow implementing solutions where the tank space is scanned automatically by the drone, and critical areas are automatically identified for closer inspection by the surveyor.

The goal of this task is to develop an algorithm for detecting structural anomalies in the video stream and correlate these with the position of the drone and the camera view direction in order to establish the location(s) of the faults.

  1. Implement the algorithm as a post-processing for detecting structural anomalies based on example survey video and in addition to anomaly categories (e.g. good, fair or poor coating)
  2. Implement the algorithm for real-time detecting structural anomalies during video based survey
Publisert 28. juni 2016 10:50


Omfang (studiepoeng)