To maintain coastal cleanliness, the Environmental Management Administration (EMA) of the Ministry of Environment in 2024 utilized unmanned aerial vehicles (UAVs) combined with artificial intelligence (AI) image recognition technology to inspect 15 coastal sites, covering a total area of 13.183 square kilometers. This initiative successfully identified, reported, and cleared over 582 metric tons of various types of waste, marking a significant breakthrough in the interpretation and application of coastal waste data.

By incorporating object segmentation into the AI recognition model, the accuracy of area interpretation improved by approximately 50%, with consistency between AI-based and manual identification exceeding 80%. Using UAVs for wide-area aerial surveys enables rapid detection of coastal pollution, facilitates real-time reporting and cleanup, and helps evaluate coastal cleanliness — restoring a pristine environment for the public to enjoy.

EMA pointed out that Taiwan has an extensive 1,990-kilometer coastline, and due to topographical challenges, relying solely on manpower for inspections is impractical. The adoption of UAV aerial photography allows each flight to accurately cover up to 2 kilometers of coastline. When paired with AI image recognition, aerial images are stitched together and analyzed for debris coverage. Locations with high pollution levels are then flagged for immediate action — all within a day. In contrast, previous methods required about a week per coastal zone for manual inspection and analysis.

In 2024, the EMA conducted 15 drone missions, capturing 5,726 aerial images, which were then stitched and analyzed to identify 477,630 square meters of waste-covered area with an AI recognition rate of over 80%. Additionally, 22 instances of coastal pollution were reported, leading to the removal of 582 metric tons of waste, significantly contributing to the protection of Taiwan’s coastal environment.

Since 2023, EMA has worked with domestic research institutions to develop AI algorithms specifically for coastal waste recognition, using over 8,000 aerial images for model training. As a result, the AI system now achieves over 80% consistency with human interpretation. This enables effective automatic detection, statistical analysis, and pattern recognition of waste accumulation, forming a solid basis for waste reduction strategies and coastal cleanliness monitoring. Looking ahead, EMA will continue enhancing AI capabilities, focusing on classifying different waste types visible in aerial images to better inform waste management practices. UAV-based monitoring will be institutionalized as part of key performance indicators, and efforts will be aligned across ministries and local governments to collectively improve coastal cleanliness.

This technological innovation not only increases the efficiency of coastline inspections and cleanup but also marks a significant step toward smart environmental governance and sustainable development for Taiwan.