A New Approach of Detection and Recognition for Artificial Landmarks from Noisy Acoustic Images

Abstract

This paper presents a framework for underwater object detection and recognition using acoustic image from an imaging sonar. It is difficult to get a stable acoustic image from any type object because of characteristic of ultrasonic wave. To overcome the difficulties, the framework consists of the selection of candidate, the recognition, and tracking of identified object. In selection of candidate phase, we select candidate as possible objects using an initial image processing and get rid of noise or discontinuous object using a probability based method in series of images. The selected candidate is processed in adaptive local image processing and recognition using shape matrix recognition method. Identified object in previous phase is tracked without selection of candidate, and recognition phase. We perform two simple tests for the verification of each phase and whole framework operability. overcome the difficulties, the framework consists of the selection of candidate, the recognition, and tracking of identified object. In selection of candidate phase, we select candidate as possible objects using an initial image processing and get rid of noise or discontinuous object using a probability based method in series of images. The selected candidate is processed in adaptive local image processing and recognition using shape matrix recognition method. Identified object in previous phase is tracked without selection of candidate, and recognition phase. We perform two simple tests for the verification of each phase and whole framework operability.