The Executive Yuan launched the "Respect the Sea" policy in 2020, aiming to establish and define coastal cleanliness. It focuses on five aspects: "cleaning, reducing, disposal, transparency, and education." The coordination among various government departments and local authorities is crucial to ensure that every inch of the coastline in Taiwan is responsibly cleaned, managed, and maintained. Currently, environmental inspections of coasts and beaches by relevant authorities and international environmental organizations rely heavily on field visits by inspection personnel, with rapid screening investigations as the primary method. This project emphasizes the integration of environmental inspection management with emerging technology – Unmanned Aerial Vehicles (UAVs) – to develop an intelligent management mechanism with operational efficiency and real-time inspection capabilities. The project conducted 10 aerial survey operations, responding to a total of 9 reports of coastal pollution, with a cleanup volume of 456.438 tons. The ortho-image analysis results showed that the waste area accounts for approximately 3.42%, a slight decrease from the 3.85% reported in the previous year's project. Due to the locations predominantly in the southwest, where aquaculture activities are prevalent, the waste types with higher proportions are mainly bamboo rafts/bamboo poles (waste oyster racks and oyster rack floats), comprising over 80% of the waste types. The project collected 36 sets of coastal aerial image materials, providing them to machine image recognition models for learning. The accuracy of interpretation compared to manual selection by personnel reached 78%. The development of AI image recognition functionality involved continuous shooting and mutually overlapping photos of coastal environments. The system automatically performed geometric correction and image stitching upon upload, presenting the waste distribution along the entire coastline. Using this image, AI recognition of waste can reduce the manual inspection time for a single mission by 2 hours, and the image stitching and manual waste digitization operation by 5 days. The project compiled information on coastal aerial survey operations conducted by local environmental protection bureaus in 19 coastal counties and cities. Two local trial operations were held at White Sands Bay in New Taipei City and Baiyu Beach in Taoyuan City. Assistance from cloud IoT and smart city initiatives led to winning the "Cloud IoT Innovation Award" Excellent Application Award. The project also adjusted the operational procedures for "Coastal Environmental Aerial Inspection" based on the specifications required for AI image recognition.

Coastal Waste Survey, Aerial Photography, Unmanned Aerial Vehicle(UAV), Orthophotos, AI Image Recognition