RiverSnap

In the RiverSnap research project (part of the Future Lab Water project), smartphone images and other sources such as drone images of rivers, as well as videos voluntarily provided by citizens, scientists and community members, are combined with in-situ data (e.g. data from official water level stations). This data is analysed to determine relevant water-related, hydraulic and structural parameters of river systems. The RiverSnap project focuses on the development and research of innovative methods for determining hydraulic parameters, which are shared on this platform.

To achieve this goal, state-of-the-art methods of machine learning and artificial intelligence are used, developed and improved. These methods enable the precise determination of various hydraulic parameters such as water level, river/ditch condition (weeds/vegetation), river course (erosion/deposition), pollution (rubbish, log jams) or water quality (turbidity). The findings are then passed on to various user groups to enable informed decisions and effective resource management.

1. Take photos of the water body (without zoom, without filter) at a fixed station (see Fig. 1 and Fig. 2) or take your own photos at a location of your choice. If possible, please activate the GPS location tag for images in your smartphone camera settings or provide us with information about the location of the photo. 
As we need a series of images from different days, please ensure that the photos are taken from the same location and a similar position/angle if possible.

2. Upload the images via direct link, share the images by email (riversnap.de@gmail.com) or via Facebook, Twitter orInstagram with a unique hashtag (#RiverSnap, #RiverSnaphannover, #LufiRiverSnaphannover, etc.).

Another option is to upload images to the Crowdwater platform. Crowdwater is a citizen science project that has established an observation network for inland waters. It also uses infrastructure such as Hany frames (CoastSnaps) to collect smartphone images.

RiverSnap Objectives:

Artificial intelligence (AI)-based parameter extraction:

• The aim of RiverSnap is to develop robust, intelligent, cost-effective and efficient AI-based analysis approaches.
• These approaches aim to extract/monitor hydraulic parameters such as water level (water level), river course (erosion/subsidence), ecological characteristics or flood conditions from the collected images/videos from various sources.

Images collected from various sources:

• Scientists, community members and citizens can participate by taking pictures or videos of water bodies with their smartphones/cameras at an official measuring station.
• Another option is to take photos at any location. To make the best use of these images, repeated shots at the same location are desirable.

Initial insights into novel image analysis methods investigated in RiverSnap:

The accurate segmentation of images from the background is crucial for a variety of tasks. Numerous segmentation methods have been developed for this task. However, many of them, especially traditional methods, face challenges due to the dynamic nature of water, its varying colours and the reflection of structures on the surface.

Machine learning/AI methods are particularly suitable for these tasks. The segmentation algorithms used for this purpose, such as the SAM algorithm (Meta AI), are trained with specific data/images to improve their parameters and accuracy (see Figure 2). Read a new publication on this topic, which deals with the comparison and improvement of various segmentation algorithms.