The aim of this internship is to develop a monitoring system for estimating the water level in rivers and water flows based on the automatic analysis of image acquisitions. The internship will take place in GIPSA-lab (Grenoble) in collaboration with EDF.
EDF-DTG (Integrated engineering unit of Électricité de France) manages a network of about 300 gauging stations on water streams located in different parts of the national territory for monitoring water resources used for power generation. A hydrometric station is composed of a staff gauge (calibrated scale for visual measurements) and a pressure sensor which continuously measures in real time the water level, as shown in Figure 1(a). However, pressure sensors are in contact with the water streams and they can be damaged during floods. Furthermore, in case of siltation or clogging with sediments and algae they provide imprecise measurements. To overcome the limitations of traditional measurement devices, since 2014 EDF-DTG has installed in some hydrometric stations cameras acquiring images with hourly frequency (examples in Figure 1(b)). These images are already used operationally to validate the measurements of the pressure sensors by human reading (check by visual inspection the water
level on the gauge).
The aim of this internship is to develop an algorithm for estimating automatically the water level from the image acquisitions. Despite the apparent simplicity of this task, images show an extremely high variability due to different environmental conditions (i.e., different illumination due to variable weather conditions, time of the day, presence of shadows, occlusions of the gauge – see Figure 1(c) for some examples) which should be properly handled by the image analysis algorithm. A possible workflow for the algorithm for the estimation of the water level is the following: i) image registration and ortho-projection; ii) image enhancement and feature extraction with image processing tools; iii) estimation of the water level via statistical inference given the extracted features.
In order to increase the precision of the water level estimations and the robustness to different environmental conditions the acquisition parameters can be optimized (e.g., use a sequence of images for each estimation). Furthermore, some sites are also equipped with a thermal camera. The joint analysis of images in the visible and thermic domain can be considered. Some of the camera installations (e.g., Voiron station) are accessible during the internship in order to test different acquisition set-ups.
(c) GdR 720 ISIS - CNRS - 2011-2018.