Photogrammetric water level determination using smartphone technology

Abstract

Rapid technological progress has made mobile devices increasingly valuable for scientific research. This paper outlines a versatile camera‐based water gauging method, implemented on smartphones, which is usable almost anywhere if 3D data is available at the targeted river section. After analysing smartphone images to detect the present water line, the image data is transferred into object space. Using the exterior orientation acquired by smartphone sensor fusion, a synthetic image originating from the 3D data is rendered that represents the local situation. Performing image‐to‐geometry registration using the true smartphone camera image and the rendered synthetic image, image parameters are refined by space resection. Moreover, the water line is transferred into object space by means of the underlying 3D information. The algorithm is implemented in the smartphone application “Open Water Levels”, which can be used on both high‐end and low‐cost devices. In a comprehensive investigation, the methodology is evaluated, demonstrating both its potential and remaining issues.

Assessing the Influence of Temperature Changes on the Geometric Stability of Smartphone- and Raspberry Pi Cameras

“…In view of the above-mentioned low-cost early warning flood systems using mobile and stationary cameras, these errors should be detected and compensated in applications of long-term observations using fixed object points in river's environment to determine the prevalent camera geometry at the time of data acquisition, e.g., using image-to-geometry registration [4,41,42]. Otherwise, image drifts can lead to false measurements of the water level, for instance considering the case that the water level increases but an image shift towards the riverbed would compensate the trend.…”

“…Otherwise, image drifts can lead to false measurements of the water level, for instance considering the case that the water level increases but an image shift towards the riverbed would compensate the trend. In case of flood observation, the errors resulting from temperature-related changes in the IOP can be neglected because the reliability of water level estimation, e.g., with methods introduced by e.g., [4,43,44], decreases strongly due to large waves at the water surface that impede a unique detection of the shore line.…”

“…Therefore, they are already a centrepiece in several early warning systems that are supported by volunteered geographic information with user-generated content [2,3]. Recently published water-level monitoring and flood-forecasting tools adapt well-established photogrammetric methods to smartphone-and Raspberry Pi (RPi) cameras to use them as versatile measurement instruments, e.g., [4][5][6][7][8]. To restore the collinearity between the 2D image and the related 3D object scene, i.e., to determine the linear relationships of 2D image points and 3D object points that lie on image rays with a shared origin called projection centre, knowledge about the interior orientation parameters (IOP) is required, which can be determined via photogrammetric camera calibration, e.g., [9][10][11].…”

“…Referring to the camera design, changes of the camera temperature may have strong impact on the camera geometry stability and thus on the measurement accuracies compared to e.g., DSLR cameras. With regard to water-level monitoring applications described by [4,5], deviations in the IOP will cause errors in the translation of 2D water lines measured in images to 3D object space. These errors range from a few millimetres to several centimetres depending on the camera-to-object distance.…”

Sensor fusion

“…Elias et al. (2019) provide an example of incorporating smartphone photogrammetry into a sensor‐fusion solution.…”