Publication: River feature morphology using k-means clustering in image segmentation of uav imagery
Date
2022-06-01
Authors
Iftekar, Ansari Emaad
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In Malaysia, flood disaster is considered to be an annual catastrophic disaster due to their consistent occurrence over the years. In this regard, flood hazard assessment models play a significant role, as they form the central component of the flood risk analysis system. With the expeditious evolution of computer techniques, processing of satellite and unmanned aerial vehicle (UAV) images for river hydromorphological feature detection and flood management have gathered pace in the last two decades. Different image processing algorithms and artificial neural networks were implemented in past studies for the monitoring and classification of river features. This study presents the application of the K-means image segmentation algorithm with image thresholding to quantify variation in river surface flow areas and vegetation growth along Kerian River, Perak, Malaysia. The river characteristic
recognition directly or indirectly assisted in studying river behaviour and flood monitoring. Dice similarity coefficient (DSC) and Jaccard index were numerated between thresholded images that were clustered using the K-means image segmentation algorithm and segmented images. Based on the quantitative evaluation, a Dice similarity coefficient and Jaccard index of up to 97.86% and 94.36% were yielded for flow area and vegetation calculation. Thus, the proposed technique was functional in evaluating river characteristics with reduced errors. With minimum errors, the proposed technique can be utilized for quantifying agricultural areas and urban areas around the river basin. Regression analysis of suspended sediment concentration and Hue Saturation Value (HSV) color space components of UAV captured river surface images were also performed in the proposed study. It was
concluded using various statistical test that the correlation between the suspended sediment concentration and HSV components of UAV captured river water surface images were non-linear. Furthermore, non-linear correlation analysis would be needed in future for obtaining an accurate relationship between the suspended sediment concentration and HSV components of aerial images.