Image side-scan sonar segmentation for seafloor detection in Lembeh strait, Bitung, North Sulawesi

Abstract

Segmentation is the process of dividing and grouping the pixel of an image into a segments or objects with the similar characteristics into homogenous areas. This method is expected to get better result with relatively faster result compared to manual (digitation) method. Side-scan Sonar image segmentation using Multi-resolution segmentation algorithm with considered many parameter and using pixel based threshold. The segment grouping using spectral reference by emphasizing the degree of brightness and texture where the two parameters are related to each. Segmentation with two parameters (brightness and texture) is effectively used for targets located near a towfish because of higher target spectral heterogeneity. TVG (Time Varried Gain) correction deliver more detailed result, image with TVG have highest accuracy of 0.987. This proves that the TVG correction that regulates the backscatter to affect the hue in the image has an important role in the seafloor classification process, in addition it can sharpen the appearance of the target which is far from the tow-fish range. Weakness of energy during measurement caused by angle and distance from instrument to target far from the towfish will affect the value of brightness and texture.