U-Net for Taiwan Shoreline Detection from SAR Images

Abstract

Climate change and global warming lead to changes in the sea level and shoreline, which pose a huge threat to island regions. Therefore, it is important to effectively detect the shoreline changes. Taiwan is a typical island, located at the junction of the East China Sea and the South China Sea in the Pacific Northwest, and is deeply affected by shoreline changes. In this research, Taiwan was selected as the study area. In this research, an efficient shoreline detection method was proposed based on the semantic segmentation U-Net model using the Sentinel-1 synthetic aperture radar (SAR) data of Taiwan island. In addition, the batch normalization (BN) module was added to the convolution layers in the U-Net architecture to further improve the generalization ability of U-Net and accelerate the training process. A self-built shoreline dataset was introduced to train the U-Net model and test its detection efficiency. The dataset consists of a total of 4,029 SAR images covering all coastal areas of Taiwan. The training samples of the dataset were annotated by morphological processing and manual inspection. The segmentation results of U-Net were then processed by edge detection and morphological postprocessing to extract the shoreline. The experimental results showed that the proposed method could achieve a satisfactory detection performance compared with the related methods using the data provided by the Ministry of the Interior of Taiwan from 2016 to 2019 for different coastal landforms in Taiwan. Within a 5-pixel difference between the detected shoreline and the ground truth data, the F1-Meaure of the proposed method exceeded 80%. In addition, the potential of this method in shoreline change detection was validated with a sandbar located on the southwestern coast of Taiwan. Finally, the entire shoreline of Taiwan has been described by the proposed approach and the detected shoreline length was close to the actual length.

Performance Evaluation and Improvement of Shoreline Detection Using Sentinel-1 SAR and DEM Data

“…According to the above definition, the dynamic changes of shoreline boundaries are easily affected by sediment movement and tidal fluctuations in the coastal zone [31]. In addition, there are some errors in SAR shoreline detection due to SAR geometric distortion, which has been shown in our previous study [29].…”

“…Therefore, it is imperative to establish a dataset containing different types of coastal landforms to develop the proposed shoreline detection method. According to our previous study [29], the labeling of the dataset was done by the Otsu's thresholding method, morphological processing and assisted by the human inspection. The Sentinel-1 images were first binarized by Otsu's thresholding method and the small textures and fragments in image segmentation results were removed through closing, opening and hole filling operations of morphological processing.…”

“…Due to some differences in the background and texture of each sub-image, the shoreline detection results are not fully connected or consistent in the overlapping images. The discontinuity of the merged shoreline was post-processed by the morphological image processing proposed in our previous study [29]. First, the closing and hole filling operations were used to connect the gaps between the detected shorelines in sub-images.…”

“…The U-Net model was modified by using the MobileNetV3 [27] backbone architecture and convolutional block attention module (CBAM) [28] to improve the feature extraction ability. Our previous study [29] showed that the geometric distortion of SAR data affected the shoreline detection performance in some regions with large slope fluctuations, especially the faulted coast in eastern Taiwan. To reduce the misclassification of these undulating regions, this study improved shoreline detection results by combining Sentinel-1 and DEM data.…”

An Effective Method of Infrared Maritime Target Enhancement and Detection with Multiple Maritime Scene

“…Infrared (IR) imaging technology has been widely used in target search and maritime navigation [1], maritime navigation [2], and maritime environmental variation [3,4]. IR imaging data can reflect object distance information and target shape information within the maritime background in two dimensions.…”