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

Zakariya, Razak; Abdullah, Mohd Azhafiz; Hasan, Rozaimi Che; Khalil, Idham

doi:10.1088/1755-1315/117/1/012047

Cited by 4 publications

(1 citation statement)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Monochromatic backscatter mosaic grayscale (high and low intensity) has been widely used as a feature in many classification techniques [4]. Several previous studies have used the MBES backscatter for bottom sediment classification [5] [12]. There are several methods used to classify bottom sediments with MBES backscatter, namely Bayesian [13] [14], neural networks [15], supervised and unsupervised methods [16] [17], combined with bathymetric data [18], visual data mining [19], combination with morphological features [20], Local Angular Response (LRA), Angular Range Analysis (ARA) [6] [21] [22] [23] [24] [25] [26] , Object Based Image Analysis (OBIA) [27], a combination of Pixel Based (PB) and Object Based (OB) [28], (least squares cubic spline approximation (LS-CSA) [7].…”

Section: Introductionmentioning

confidence: 99%

Seabed Sediment Classification using Multi-Frequency MBES Bathymetric and Its Features

Khomsin,

Mukhtasor,

Suntoyo

et al. 2023

IOP Conf. Ser.: Earth Environ. Sci.

View full text Add to dashboard Cite

Principally, for a few decades, hydrographical surveys have been only to determine the depth of seawater. Sonar measurement tehcnology is the most widely used to conduct hydrographical surveys employing a singlebeam echosounder (SBES) and multibeam echosounder (MBES). In addition to depth information, seafloor sediment distribution maps are essential for port planning and management. In general, the distribution of seafloor sediments is predicted using backscatter data from SBES and MBES at single and multiple frequencies. The bathymetric data generated by the echosounder can be derived into several bathymetric features such as slope, ruggedness, roughness, aspect, bathymetric position index (BPI), and curvature. This study examines the possibility of using bathymetric measurement and bathymetric derivation of multi-frequency MBES to predict the distribution of seafloor sediments, especially in the harbor pond area. The study used a deep neural network (DDN) to classify the distribution of seabed sediments with bathymetric and bathymetric features input, validated with 74 in situ sediment samples (clayey sand, silt, sandy silt, and silty sand). Up to 75% of data sample sediments are used for training and 25% for validation. The classification results by DNN showed 42.6% clayey sand, 7.4% sandy silt, 46.7% silt, and 3.35% silty sand. The overall accuracy (AO) and Kappa classification of seabed sediments with DDN were 59.5% and 0.54 (moderate), respectively.

show abstract