Semi-Supervised Water Boundary Detection using Drone Imagery
Water boundary detection is an essential component in coastal research, especially in the fields of ocean engineering and oceanography. Detecting the boundary between water and land on coasts gives insight into the tidal patterns of the ocean. The task has inherent difficulties due to changing weather patterns and light levels over the water, leading to different colors of water in images. Similarly, drone imagery has a large volume of data without proper labelling, resulting in significant time delays and human intervention to accomplish the task. In this paper, a semi-supervised model using drone imagery is proposed to minimize the level of human interaction and increase the accuracy of water boundary detection. Experimental analysis on k-means clustering, k-nearest neighbors classification, logistic regression, support vector machine classification, and c-support vector machine classification are presented.