Support Vector Machine and Decision Tree Based Classification of Side-Scan Sonar Mosaics Using Textural Features

Abstract

<p><strong>Abstract.</strong> The diversity and heterogeneity of coastal, estuarine and stream habitats has led to them becoming a prevalent topic for study. Woody ruins are areas of potential riverbed habitat, particularly for fish. Therefore, the mapping of those areas is of interest. However, due to the limited visibility in some river systems, satellites, airborne or other camera-based systems (passive systems) cannot be used. By contrast, sidescan sonar is a popular underwater acoustic imaging system that is capable of providing high- resolution monochromatic images of the seafloor and riverbeds. Although the study of sidescan sonar imaging using supervised classification has become a prominent research subject, the use of composite texture features in machine learning classification is still limited. This study describes an investigation of the use of texture analysis and feature extraction on side-scan sonar imagery in two supervised machine learning classifications: Support Vector Machine (SVM) and Decision Tree (DT). A combination of first- order texture and second-order texture is investigated to obtain the most appropriate texture features for the image classification. SVM, using linear and Gaussian kernels along with Decision Tree classifiers, was examined using selected texture features. The results of overall accuracy and kappa coefficient revealed that SVM using a linear kernel leads to a more promising result, with 77% overall accuracy and 0.62 kappa, than SVM using either a Gaussian kernel or Decision Tree (60% and 73% overall accuracy, and 0.39 and 0.59 kappa, respectively). However, this study has demonstrated that SVM using linear and Gaussian kernels as well as a Decision Tree makes it capable of being used in side-scan sonar image classification and riverbed habitat mapping.</p>

Assessment of a Low-Cost Side-Scan Sonar for River Estuary Underwater Imaging

“…As can be found in (Lianantonakis & Petillot, 2007;Hamilton, 2015;Buscombe, 2017;Hamill et al, 2018) who used these textures in sonar image analysis. The combination of both first and second-order also showed a promising result in sonar image classification (Febriawan et al, 2019). Those approaches can be applied using low-cost sonar image in this study.…”

“…Additionally, the development of machine learning techniques can be a prominent topic in sonar imagery classification. For instance, Doherty et al (1989) used Decision Tree (DT) in sonar classification, Rhinelander (2017) who explored sonar image classification using Support Vector Machine (SVM), and Febriawan et al, (2019) who compared both DT and SVM based classification of side-scan sonar mosaics. Other machine learning techniques such as Random Forest and Naïve Bayes also can be another option to study.…”

Examining the Links between Multi-Frequency Multibeam Backscatter Data and Sediment Grain Size

“…Different approaches have been developed for texture analysis for various applications. In the marine science domain, the grey level co-occurrence matrix (GLCM) method is widely utilized [2,7,23,27,30,40]. Basically, GLCM provides information about the number of pixel combinations that are separated by a specified distance in each direction [30].…”

“…This exploration revealed more detail compared to the use of a single statistical measure of distribution (e.g., mean), which always masks out the fine scale variability. However, this high level of detail also rendered these small images to the influence of noise and size-specific texture sensitivity, hence caution needs to be exercised while interpreting the texture intensity [10,30,40].…”

“…In contrast, the correlation between the individual texture features (contrast and correlation) and sediment grain size was generally weak and statistically insignificant (p > 0.05), suggesting that backscatter intensity is the most important predictor variable in this study. The poor performance of the GLCM texture can be linked to their sensitivity to survey geometry [30], analysis windows, distance between pixels, and the number of grey levels [10,27,40]. Similarly, the GLCM texture relies on pulse length and is unreliable at a high incidence angle [54].…”