Active Learning for Recognition of Shipwreck Target in Side-Scan Sonar Image

Abstract

In order to realize the automatic and accurate recognition of shipwreck targets in side-scan sonar (SSS) waterfall images, a pipeline that contains feature extraction, selection, and shipwreck recognition, an AdaBoost model was constructed by sample images. Shipwreck targets are detected quickly by a nonlinear matching model, and a shipwreck recognition in SSS waterfall images are given, and according to a wide set of combinations of different types of these individual procedures, the model is able to recognize the shipwrecks accurately. Firstly, two feature-extraction methods suitable for recognizing SSS shipwreck targets from natural sea bottom images were studied. In addition to these two typical features, some commonly used features were extracted and combined as comprehensive features to characterize shipwrecks from various feature spaces. Based on Independent Component Analysis (ICA), the preferred features were selected from the comprehensive features, which avoid dimension disaster and improved the correct recognition rate. Then, the Gentle AdaBoost algorithm was studied and used for constructing the shipwreck target recognition model using sample images. Finally, a shipwreck target recognition process for the SSS waterfall image was given, and the process contains shipwreck target fast detection by a nonlinear matching model and accurate recognition by the Gentle AdaBoost recognition model. The results show that the correct recognition rate of the model for the sample image is 97.44%, while the false positive rate is 3.13% and the missing detection rate is 0. This study of a measured SSS waterfall image confirms the correctness of the recognition process and model.

Automated Stone Detection on Side-Scan Sonar Mosaics Using Haar-Like Features

“…This, however, increases false-positive detections, as, e.g., the scattered noise of the mosaic might be interpreted as individual objects. Concomitantly, a detector, which was trained for larger objects, will miss smaller objects [28]. A large detector might further become sensitive for extensive transitions between sediments showing a prominent change of the acoustic backscatter (e.g., fine sand to coarse sand, cf., Figure 1d).…”

“…Hence, selection of the appropriate size for the detector implies a trade-off between the general detection rate of objects and a small amount of false-positive detections. Most importantly, a large training dataset consisting of both positive and negative images is required for optimal training and an accurate detector [28]. In particular, for special applications such as the detection of stones such a training dataset cannot as yet be obtained elsewhere like training sets for objects such as faces, cars, trees and the like (e.g., Open Images Dataset [29] and MS-COCO [30]).…”

Hydroacoustic Mapping of Geogenic Hard Substrates: Challenges and Review of German Approaches

“…In principle, these methods may be applicable to the detection of stones, although no corresponding studies exist to our knowledge. Recent examples include the use of independent component analysis to extract features of high resolution backscatter mosaics for the detection of smaller metallic objects [99] or shipwrecks [100]. Environmental conditions are included in the analysis by Williams and Fakiris [101].…”

The Effectiveness of Using a Pretrained Deep Learning Neural Networks for Object Classification in Underwater Video

“…The paper [9] tackles the problem of automatic detection and classification of underwater mines on images generated by a Synthetic Aperture Sonar (SAS) using the DLNN. In [15], the recognition model of a shipwreck target using side-scan sonar and CNN is presented. Although the sonar imagery has a lower resolution or grey-scale color, the corrected classification of sonar imagery for jellyfish detection presented in the paper [16] is improved by up to 90%.…”