This paper presents AUVs landmark system for the positioning reference. A cubic landmark and 3D model based recognition method are proposed. It can be detected by acoustic and optical sensor. The landmark is initially found using the topological characteristics. Then its relative position is estimated. Then its 3D CAD model is modified to the position and matched with the landmark in image. Experimental landmark was fabricated and tank tests were carried out to evaluate the performance of the proposed landmark.


AUVs(Autonomous Underwater Vehicles) have been successfully automated long range data collecting missions due to their excellent mobility and light deck support. Their navigation mainly has depended on acoustic gudance signal or dead reckoning. The former requires heavy deck support or installation load, the latter couldn''t reset position error by floating on surface. A pre-installed man made landmark on a route can be an alternative. They are able to serve for navigational aids such as precise absoute position updating tool. They enables the AUVs navigates with high-accuracy and to undertake complicate tasks such as underwater manipulation(Yuh, 1998) and structure safety inspection(Yu, 2004)The landmark recognition is also the inital step for a general object recognition which is essential for risky tasks such as safety inspection of underwater structures. The followings are the representative AUV''s landmark studies which succeeded in test. Maki(2006) proposed floating pole type acoustic landmark. It guarantees reliable detection in wide range but it provides limited position variables(X, Y) and the application is limited. Park(2009) proposed a lighting lamp type landmark for AUV''s docking. 5 strong lights were attached on a circular frame. They represent the landmark''s feature points. This method extends the landmark''s detecting range and improves the reliability. But it needs continuous support load; power supply and maintenance. Yu(2001) proposed colored spheres attached landmark.

This content is only available via PDF.
You can access this article if you purchase or spend a download.