Preliminary study on a framework for imaging sonar based underwater object recognition

This paper presents a framework for underwater object recognition using imaging sonar. The framework consists of selection of candidates of interest, recognition, and tracking. Instead of trying to recognize objects from a whole image at any certain time using one-size-fit-all method, we're going to select candidates as possible objects of interest first and get rid of fake candidates using a probability based method similar to particle filter in series of images. Each candidate in small cut-out image is under processing by various and specific image processing techniques to recognize object, then it is transferred to tracking phase with object ID. We perform a simple test for an artificial landmark to show feasibility of the proposed framework.