Extensive inspection of underwater structures can be very hazardous and requires a very high degree of automation in the use of AUVs (Autonomous Underwater Vehicles). In order to accomplish the autonomous inspection task, AUVs should identify and investigate an inspection target by means of image processing.
In this paper, an image recognition method is proposed for AUVs to identify an inspection target that may have been deformed by marine bio-fouling or corrosion. The proposed method incorporates improvement of the lighting method that also makes full use of underwater environment's characteristics for identification of a deformed object. This method allows robust recognition of a deformed object with small computing power available on AUVs. As an example, a recognition method for sacrificial anodes, an important inspection target, is addressed. The effectiveness of the proposed method is demonstrated by simulations and also by field experiments carried out under the bottom of Mega-Float; a very large floating artificial land located off Yokosuka.