Smartphone-Based Water-Level Estimation Method Using Visual Localization and Ray Casting

Abstract

Monitoring changes in river water levels is essential for accurate flood prediction and disaster prevention. However, existing fixed water level measurement systems are often costly and inefficient for large-scale monitoring. This study presents a smartphone-based method for estimating water levels that integrates 3D reconstruction, visual localization, image-based water body segmentation, and ray-casting techniques. This method estimates the camera’s position and pose from captured images by refining these parameters using homography and IMU data. The rays were then cast along the segmented water body contours in the images to determine the intersection points with the 3D reconstructed area of interest. The water levels were quantified with a centimeter-level resolution (average error <2-cm), demonstrating the high accuracy of the proposed system. To our knowledge, this is the first attempt to estimate water levels from a single image by first determining the camera’s position using image-based visual localization and then applying ray casting with a pre-constructed 3D model, forming a novel paradigm for scalable, infrastructurefree hydrological monitoring. Field experiments conducted at the Sebyeong Bridge in the Oncheoncheon Stream Basin in Busan, South Korea, validated the effectiveness of the system, showing its potential as a reliable and scalable solution for flood monitoring. By leveraging widely accessible smartphones, the proposed method enables broad public participation in water-level monitoring and offers a cost-effective and sustainable alternative to traditional systems. Furthermore, its adaptability to diverse river environments makes it suitable for real-time flood monitoring and lays the foundation for an integrated river management platform. While real-world performance may be affected by environmental variability, the proposed method suggests the potential of climate disaster response technologies and may form a basis for future advances in flood prevention and disaster management.