Waterline Extraction for Artificial Coast With Vision Transformers

Abstract

Accurate acquisition for the positions of the waterlines plays a critical role in coastline extraction. However, waterline extraction from high-resolution images is a very challenging task because it is easily influenced by the complex background. To fulfill the task, two types of vision transformers, segmentation transformers (SETR) and semantic segmentation transformers (SegFormer), are introduced as an early exploration of the potential of transformers for waterline extraction. To estimate the effects of the two methods, we collect the high-resolution images from the web map services, and the annotations are created manually for training and test. Through extensive experiments, transformer-based approaches achieved state-of-the-art performances for waterline extraction in the artificial coast.

Development and application of a more refined process for extracting rock crack width information based on artificial intelligence

“…Although the Transformer-based SegFormer network shows superior performance over the CNNbased network in several datasets [20,21,22], the CNN-based network still shows excellent performance for the crack image data in this study. In addition, the SS network can achieve more accurate segmentation in the dataset reconstructed based on RCAN, and the overall structure and detail aspects of the detected cracks best match the labeled data.…”

“…With the rapid development of CV technology, some novel algorithms such as FCN, Deeplab V3+ and FraSegNet are gradually being applied to the recognition and parsing of various research objects and have shown better performance in feature extraction [14,15,16,17,18,19]. The feature extracted by the latest Transformer-based algorithms such as SETR, TransUNet and SegFormer have richer global contextual information, which overcomes the drawback that convolutional neural network-based (CNN) algorithms cannot directly extract long-range information and eventually show excellent performance in semantic segmentation tasks [20,21,22]. However, in the case of regional rock crack images, the selection of hyperparameters in the feature extraction process is highly dependent on the image resolution and specific scene features.…”

DAENet: Deformable Attention Edge Network for Automatic Coastline Extraction from Satellite Imagery

“…Recently, transformers have demonstrated significant benefits in natural language processing and computer vision, exhibiting outstanding performance across numerous tasks. The first introduction of transformers in this domain was by Yang et al [42]. They first introduced transformers into this domain by experimenting with the pure Transformer architecture SETR [43], achieving performance comparable to existing CNN methods in land-sea segmentation.…”

TCUNet: A Lightweight Dual-Branch Parallel Network for Sea–Land Segmentation in Remote Sensing Images

“…Coastline information is an important basis for the implementation of coastal zone protection and disaster management, the basis for the development and use of marine resources, and an important territorial resource for countries bordering the sea, and plays an significant role in the ecological safety of the ocean [3]. However, at the same time, the extraction of the coastline is a very challenging problem, because it is the land-water boundary of the multi-year average high tide, rather than an instantaneous line [4]. Traditional shoreline extraction methods are mainly manual measurements.…”

“…Thus, the detected edges are not accurate enough. At the same time, these methods are less efficient, and can only detect the significant edges in the image, and the accuracy of the obtained boundaries is not high [4,15,16]. The object-oriented classification method combines pixels into objects, integrating their interrelationships and spatial distributions, and thereby reducing the interference from internal pixel information, and maximizing the utilization of image information.…”