Visual inspection of hydroelectric dams using an autonomous underwater vehicle

This paper presents an automated solution to the visual inspection problem of hydroelectric dams. A small autonomous underwater vehicle, controllable in four degrees of freedom (surge, sway, heave, and yaw), is used for autonomously exploring the wall of a dam. The robot is easily programmed using a mission control language. Missions are executed by an intelligent control architecture that guides the robot to follow a predefined path surveying the wall. During the mission, the robot gathers onboard navigation data synchronized with optical imagery, sonar, and absolute navigation data, obtained from a moored buoy equipped with an ultra‐short baseline system. After the mission, the images of the wall are used to build a photomosaic of the inspected area. First, image features are matched over the image sequence. Then, navigation data and interimage correspondences are optimized together using bundle adjustment techniques. Thus, a georeferenced globally aligned set of images is obtained. Finally, a blending algorithm is used to obtain smooth seam transitions among the different images that constitute the mosaic, compensating for light artifacts and improving the visual perception of the scene.