Heart of the project is the monitoring of a complex flow situation along a 110° river bend at Eggrankkurve, Thur River, in the framework of revitalization works. Therefore, we are developing and testing a low-cost airborne velocimetry system to measure large scale surface velocity fields. The measurement equipment consists of an ultra-light actioncam and a ready-to-fly lowcost quadrocopter (Fig. 1).
Video recordings are performed from heights between 40 to 90 m covering a total reach length of 300 m, while spruce chips with an edge length of 60 mm are added as tracer particles. Challenging for the image analysis are unsteady camera shakes due to the inevitable non-stable position of the Unmanned Aerial Vehicle (UAV). Each lens-corrected frame is therefore automatically orthorectified to riparian ground reference points as detected by computer vision techniques. The positional error of each point is computed to 0.15–0.40 m, so that the magnitude of the related descaling error is below ±2%, and the error of apparent ground velocity is approximately 0.03 m s-1 (maxima typically at 0.05–0.10 m s-1). These values describe the uncertainty added to the subsequently calculated particle image velocity field (Fig. 2). Over the entire reach length the final raster resolution is 1.0×1.0 m2 with 50% overlap. A comparison to velocity profiles measured simultaneously by a 3D acoustic Doppler current profiler indicates that our new type of velocimetry system is capable to describe time-averaged surface flow fields with high accuracy.
Automatic object detection to analyze the geometry of gravel grains – a free stand-alone tool
Within this project, an automated procedure is developed to estimate the grain size distribution of a gravel bed by analyzing a digital top-view photograph of it. The MATLAB based methodology is inspired by an approach originally developed at Loughborough University, but optimized to obtain a more precise separation in different area elements referring to top view areas of single grains. Moreover, a graphical user interface will enable semi-automatic control and optional correction of the detected elements.
The surface area of each grain is replaced with an ellipse of the same normalized second central moments. Fig. 1 shows an exemplary result of the acutal test version of the optimized VAW-intern MATLAB-Code. Here, each area detected as a single grain-element is color-coded differently. The main axes are highlighted in white. The grain-by-number statistic of the minor axes lengths is transformed by Fehr's (1987) method to obtain a quasi-grain size distribution in volume percent – a depiction as typically given by a classical laboratory sieve analysis - in opposite to a grain-by-number statistic as gained originally from the object detection. The grain size distribution of the subsurface layer is determined by using an empirical estimation of the percentage of non-detected finer grains. As shown in fig. 2, the grain size distributions obtained by automatic object detection ('AO', in blue), in-situ line sampling analysis ('LZA', in red) and laboratory sieving of a volume sampling ('VP', in black) are in a good agreement concerning the essential geometric parameters.
The time effort for a grain size analysis by automatic object detection is only at a fraction of the time needed for classical methods. A further benefit is that additional parameters can be provided for each grain such as: ratio of minor axis/major axis, area, perimeter, center coordinates, and the grain orientation in a horizontal plane. Factors which may decrease the quality of the results are: cast shadowing, high percentage of fine material, partly wetted or partly covered stones, and intra-granular textures.
Free access to the newly developed stand-alone tool named BASEGRAIN 1.0 including a graphical user interface (GUI) is given since 2012/08/10.
The objective of this study is to analyze the development from a constructed channel towards a nature-oriented river-bed as initialized by small scale interventions. This work will be done on the basis of a short reach of the River Töss near Winterthur (CH). The River Töss that has been corrected to a channel mainly with respect to flood protection purposes since the 1880s. Due to this ’correction’ the river developed a deficit in bed load, and thus emerging erosion processes were stopped by installing sill beams. In 2001, a first measure were started to re-initialize the self dynamisation of the river bed. Bank protections and water management structures were removed on a length of about 300 m, and an artificial island was installed in the middle of the river to lead the flow to the banks (fig. 1). However, this attempt failed as the banks stayed stable.
In 2010, an advanced change was done to initialize this self dynamization more effectively (fig. 2). This project has a pilot character for initiate self dynamization of rivers by a small scale intervention – in opposite to so far performed fluvial revitalizations realized by intensive mechanical operations to reform a whole river reach. Therefore, the VAW supervises this project scientifically and monitors the development of bed material composition, hydraulics, sediment transport and bed morphology. Special focus is given to grain size analyzing by image processing and the use of hydro-numerical means for predicting morphodynamic aspects. In the end, guidelines to initiate the self dynamization of rivers will be given to increase the effectiveness in design and performance of comparable river vitalization projects.
The monitoring is conducted since summer 2009 up to the end of 2015
Supervision and granulometric analysis of photo-optical underwater-sampling of river beds
Within this project a procedure is developed to classify the grain-sizes of a river bed by contact-free measurements at underwater conditions. A VAW in-house MATLAB code for automatic object detection to estimate the quasi-grain size distribution from top-view photographs of dry and free accessable alpine and prealpine river banks is adapted to be used in this project.
Customary methods to analyze the grain structure of a river bed need either extensive extraction, transport and laboratory sieving of the bed material or time consuming in-situ counting of single grains (e.g., line-by-number or grid-by-number sampling). These methods need free access to the bed, i.e. the river bed should not be overflowed. Otherwise the sampling is impossible in most of the cases. For larger rivers like the River Rhine, the access to the bed along the thalweg can be given by diver bells from special ships, e.g. the MV Carl-Straat. Beside the high technical efforts, an essential disadvantage of this method is that the water in the diver bell needs to be the blown out before entering it, which leads to an unspecified loss of fines. Furthermore the maximal volume per sample is limited to a few liters. Consequently, the representativity of the taken sample cannot be confirmed. Another technical device used by BAW to monitor the subaquaeous bed is a video-optical scanner consisting of an underwater camera mounted in a streamlined housing. The construction, called 'fish', is lowered to the river bed by a heaver. However, due to the low recording quality the data sets obtained up to know are not adequate enough for a systematic analysis of the grain size distributions of the river bed materials.
The project has just started. First photographs were taken within a diver bell at overpressure conditions at the bed of the River Rhine. These pictures will be analyzed using automatic object detection and the results will be compared with the grain-size distributions of the parallel taken scrape-samples. As an option, ongoing phase photographs are planned to be taken by divers and by an optimized underwater robot. These data will then be evaluated with respect to the image quality and thus its effect on the results. A successful development of a contact-free and non-destructive method to analyse the underwater grain-size distribution can be considered as a quantum leap in river bed investigations and in the fields of offshore prospection as well.