Poster Session Information

Keywords

• Rivers
• Image processing
• Aerial video

Outline

Monitoring river conditions and identifying problems are essential to preventing natural disasters such as flooding. Automating these processes is expected to significantly reduce disaster risks. This study proposes a method that combines drone-captured video with image processing techniques to enable safe, wide-ranging, and cost-effective river monitoring. The aim is to establish foundational technology for automating anomaly detection.
The proposed approach combines color detection in the HSV color space with centroid calculations across video frames to reliably identify river regions. It also incorporates a flood-filling process to evaluate pixel value differences in neighboring areas, improving detection accuracy. Together, these techniques enable efficient and consistent identification of river areas, even under varying water colors and lighting conditions.
This method enhances the accuracy and reliability of river region detection and provides a robust foundation for automating the detection of problems with rivers. By performing river area detection using moving images, this approach has the potential to contribute to research areas such as measuring flow velocity and detecting suspended matter—applications that cannot be achieved with still images. The proposed method is expected to play a critical role in disaster response and river management, reducing risks and enabling faster, more effective actions in real-world situations.