Underwater Acoustic Feature Extraction Based on Bidimensional Empirical Mode Decomposition in Shadow Field

Abstract

Recent developments in feature extraction based on bidimensional empirical mode decomposition (BEMD) is mainly about the optical image. Here we applied the BEMD to the underwater acoustic image feature extraction. Acoustic shadow fields always disturb feature extraction. To reduce it, we decomposed the underwater acoustic image into several Intrinsic mode functions (IMFs) and a residue. Thus, the Canny edge detector could extract better features from the first IMF. Sometimes, it is necessary to detect the sum of the first two IMFs. Experiments prove that this novel method really enhance the features of objects (physiognomy and texture) in the acoustic shadow field and weaken the edge of the acoustic shadow field.