| Popis: |
The gauging of open channel flows in waterways provides the foundation to monitor, understand and manage the water resources of our built and natural environment. Several methods are available for measuring the flow, with each of these methods having its own advantages and limitations. For a significant economic and environmental cost, hydraulic control structures can be built to measure the flow using analytical relationships with water height often by measuring the pressure head invasively in the water. Another common approach using the proxy measurement of water height without a hydraulic control structure is the expensive development and maintenance of a discharge rating table relating the measured water height to an estimated flow which has been manually measured at a previous time by acoustic instruments with technically proficient operators. Whilst these approaches are typically able to reasonably estimate flow within their measurement range, the safety risks in monitoring high flow events and the ongoing costs involved are prohibitive to increasing the spatial coverage of these approaches. As water resources become increasingly vulnerable to climate variability, modification of waterways, and increased extraction, there is a critical need to develop monitoring tools that can be flexible, cost-effective, and safe.Much research has been undertaken into optical non-contact methods to estimate flow in waterways by measuring surface velocities without intrusive instruments or structures. However, to date, these surface velocimetry methods are limited to a narrow operational window of certain stream types and flow velocities due to inherent challenging optical variability in stream environments. A cost-effective stereographic camera-based stream gauging device has been developed for rapid stream gauging through the remote sensing of water height and stream velocities to estimate flows and employ the learning of an adaptive discharge rating envelope. The device includes embedded edge computing capabilities, local app connectivity for setup, and online cloud fleet management with a data dashboard for streamlined deployment and ongoing operational monitoring. Automated analysis is performed reconstructing the point cloud of the scene in front of the camera out to 40 m in order to estimate the water level without any instream equipment. An optical flow algorithm is passed over the short videos collected, generating an array of net motion in the scene which is projected out of the image plane onto the assumed water surface plane using the water level estimation combined with the accelerometer and the embedded intrinsic camera properties. The optically measured motions which are out of the plane of the waterway surface are then able to be automatically filtered and integrated into a water level indexed learning surface velocity distribution which generates an updating adaptive discharge rating envelope for the site. With over 100,000 videos recorded and analysed across 20 sites, the computer vision stream gauging approach has achieved discharge measurements within 15% RMSE of traditional acoustic gauging. This work evaluates this innovative approach across sites on the east coast of Australia and demonstrates the potential to improve the operational reliability and performance of surface velocimetry stream gauging. |
