Automatic classification techniques for type of sediment map from multibeam sonar data

3rd International Conference on Research Methodology for Built Environment and Engineering 2017

生态环境科学

Zakariya, R.^1 ; Abdullah, M.A.^1 ; Hasan, R Che^2 ; Khalil, I.^1

School of Marine and Environmental Sciences, University Malaysia Terengganu, Terengganu, Malaysia^1

UTM Razak School of Engineering and Advanced Technology, University Teknologi Malaysia (UTM), Jalan Semarak, Kuala Lumpur

54100, Malaysia^2

关键词: Automatic classification;    Backscatter strength;    Ground truth data;    Kappa coefficient;    Multi-beam sonar data;    Multibeam echo sounder;    Overall accuracies;    Sediment samples;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/117/1/012047/pdf
DOI  :  10.1088/1755-1315/117/1/012047

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

Sediment map can be important information for various applications such as oil drilling, environmental and pollution study. A study on sediment mapping was conducted at a natural reef (rock) in Pulau Payar using Sound Navigation and Ranging (SONAR) technology which is Multibeam Echosounder R2-Sonic. This study aims to determine sediment type by obtaining backscatter and bathymetry data from multibeam echosounder. Ground truth data were used to verify the classification produced. The method used to analyze ground truth samples consists of particle size analysis (PSA) and dry sieving methods. Different analysis being carried out due to different sizes of sediment sample obtained. The smaller size was analyzed using PSA with the brand CILAS while bigger size sediment was analyzed using sieve. For multibeam, data acquisition includes backscatter strength and bathymetry data were processed using QINSy, Qimera, and ArcGIS. This study shows the capability of multibeam data to differentiate the four types of sediments which are i) very coarse sand, ii) coarse sand, iii) very coarse silt and coarse silt. The accuracy was reported as 92.31% overall accuracy and 0.88 kappa coefficient.