Hessel Winsemius
Hessel is trained as a hydrologist. He has gained about 20 years of experience with flood and drought risk modelling and understanding, as well as flood and drought monitoring and forecasting. Much of his work at Deltares focused on global to local flood risk management for national institutes, clients in International Finance such as the World Bank, and the insurance sector. He has authored and co-authored about 80 papers. His approach to development of technology and projects is strongly anchored in co-design. Over the past 5 years, his focus has been on research and development of locally sustainable technology and methods for monitoring related to natural hazard risks. He creates a bridge between local users and entrepreneurs with skills to operationally collect, use and operationalize data, and local and international audiences that require such data, such as National Hydrometeorological Service Providers, water-related consultancies, International Financing Institutes and Earth Observation centres. Within Rainbow Sensing, as a self-trained developer, he conceived and developed the idea for OpenRiverCam, the first entirely free and open-source software ecosystem for operational management of river monitoring sites using camera observations. Within Delft University and IHE Delft, he supervises students that work on affordable or repurposed technology for environmental monitoring. He works for and/or collaborates with internationally recognized entities such as Deltares, Rijkswaterstaat, The Netherlands Enterprise Agency, the World Bank, Humanitarian OpenStreetMap Team and GroundTruth SA.
Sessies
Background: OpenRiverCam is a free and open-source ecosystem for measuring water levels, velocities and discharge using image-based methods. The main input is a short video of the water surface. Videos are geographically registered, preprocessed, orthorectified, analyzed for velocities, and combined with cross-section information to yield water levels and flows. It converts pixel distances into meters, extracts moving details, uses cross correlation to estimate velocities, and combines pre-measured bathymetric surveys to yield water levels and discharge. For a very clear example, check out the front page of https://openrivercam.org. The first two software products created in this ecosystem provide fully operational, autonomous and automated measurements for applications such as irrigation, real-time hydropower control, flood early warning and infrastructure design:
OpenRiverCam Operating System (ORC-OS): runs on a SBC in the field and operationally collects and processes videos. Analyses can be sent to the cloud if the user wants this. ORC-OS is equipped with a user-friendly web interface for configuration, live monitoring, and direct access to data and videos stored on the device.
LiveOpenRiverCam (LiveORC): a REST API receiving videos from ORC-OS devices, centrally storing these and serving these out as raw downloads or to applications for e.g. decision support.
The cool thing for this workshop: all ORC applications are built on a Python API with command-line interface called pyOpenRiverCam (pyORC). PyORC makes available all methods through API including image stabilization, pre-processing, orthorectification, velocimetry analysis, discharge estimation and intuitive plotting. It also offers workflows for command-line processing of videos into water levels, velocities, and discharge. Moreover, if image registration control points are geographical, all outputs are geographical as well! Data can be exported to QGIS compatible meshes, and visualized, and trace simulations and combinations with other geographical data can be made.
In this 90-minute workshop we will install pyORC and process one or two drone videos into velocity products that can be ingested in QGIS. We will go through the entire process up to the point where you see data in QGIS. Bring your own laptop and make sure QGIS and Python are pre-installed! Linux or Windows? Both is fine!
Over a decade ago, researchers started to investigate how to monitor river flows with camera video footage. This offers the promise that river monitoring may become low-cost, and carried out with off the shelf IoT materials. This led to first successful applications of river flow monitoring from videos. The workflow typically consists of acquiring a video from an oblique angle, pre-process to bring forward moving tracers such as advecting ripples, bubbles or eddies, orthorectification (i.e. reprojecting pixels to meter distance, some form of tracing patterns or particles from frame to frame and from location to location, and combining velocities with a surveyed cross-section to yield river flow in cubic metre per second. Several free and sometimes open-source existing applications allow users to load videos in a user interface, and perform the steps interactively. However, operational, real-time, automated flow observations with camera videos so far was only feasible with proprietary hardware and software solutions, compromising the promise of affordability and local maintenance of services.
With OpenRiverCam, for the first time any user can acquire materials for a station, setup an automated workflow using the on-site OpenRiverCam Operating System (ORC-OS), and even collect data centrally on a central server-side REST API called LiveOpenRiverCam (LiveORC). Not only does the ecosystem provide software, it also provides extensive doumentation that guides a user how to establish a field hardware setup, and how to install and survey the station. The documentation forms an integral part of the ecosystem, as it allows a user truly become independent of the developers. In this presentation, I will demonstrate the current capabilities of the software, the underlying methods, the on-site application ORC-OS and how it runs on a standard Raspberry Pi 5, and the REST API LiveORC. I will also demonstrate how first users are able to establish services in their own local context with locally fit business propositions with examples from Zambia, California – US, and Indonesia.