dc.description.abstract | With the unique geological setting, Taiwan Strait was formed by shallow bathymetry and gentle topography composed of sandy substrate types. The depth of this area seldom exceeds 100 m, and it is shallower than 20 m in the Taiwan Shoal area. Recently, the abundant sea sand resources contained in this area have also been coveted by many illegal businesses. Therefore, in recent years, there have been frequent cases of illegal cross-border sea sand mining around the Taiwan Strait, especially in Taiwan Shoal, Kinmen Islands, and Matsu Islands. Apart from destroying marine ecology, the greatest problem of illegal sand mining is the consequential retreat of the neighboring coastline.
To address this problem, the objective of this research aims to take advantage of Synthetic Aperture Radar (SAR) technology in satellite remote sensing and to monitor the spatial-temporal hot spots of unidentified vessels in specific areas. Spaceborne SAR instruments have the advantages of superior penetration, high resolution, and independence from sunlight, making them a great tool for ocean object detection. This research uses Sentinel-1 SAR imagery as the data source. Three image frames covering the Taiwan strait are chosen to observe three main target areas with a higher number of illegal sand mining cases in recent years. Images from 2017 to 2021 are obtained to analyze the temporal variation of the hot spot. A detection system has been developed for sand carrier and sand dredger detection with SAR imagery. The workflow of this detection system is mainly composed of four steps, including preprocessing, land masking, prescreening, and discrimination & classification. Our preliminary results show that the developed algorithm can automatically detect targets over a specific size. The detection system was tested on 15 randomly selected images and the results were validated with manually identified ground truth data. The validation result achieved a detection accuracy, in terms of the F1-score, at 0.88 on average. The spatial-temporal hot spot analysis shows different movements of hot spots between these three target areas while the detection number peaks around 2019 and 2020. It is concluded that Sentinel-1 SAR images could serve as a tool for ship detection, and the detection system is able to monitor the distribution of unidentified ships in specific area, which is conducive to improving the efficiency of follow-up law enforcement.
| en_US |