Unmanned aerial vehicle (UAV) based data ascertainment of riverine systems is strongly increasing, as it is a cost effective way to gain data. It be-came state of the art in various remote sensing applications in diverse engineering fields over the last years. In this work we present a study of techniques for grain size determination of the top layer above and under the water surface, using UAV based aerial images. The results are compared with different conventional methods, such as sieving and pebble counts. The proposed methods enable the creation of detailed substrate maps, based on high precision orthomosaics with reference grain size distribution as an input for various modeling applications.

The study area is at the River Jachen, which is a 23 km long river in South Germany. The Jachen is the natural outflow of Lake Walchensee, with an altered flow regime from hydropower production. The only hydraulic structure in the river is a weir, approx. 1.5 km above the outlet of River Jachen into River Isar. The structure divides the river in an in-take ofahydropower plant and the old riverbed, supplied with a minimum flow. A custom made hex copter with a DJI NAZA-M V2 Multirotor Autopilot is used for the data ascertainment. The maximum payload of the hexacopter is approx. 2 kg. The hexa-copter carries a Sony Nex 6 camera with a 16 Mega-pixel APX-C Sensor with a 30 mm fixed focal length lens. In addition a DJI Phantom 3 Professional quad-copter was used for further aerial images. The quad-copter carries an integrated camera with a 12 Mega-pixel Sony EXMOR 1/2.3” Sensor and a 20 mm fixed focal length lens on a on a 3 axis. The maximum flight time of the setup is approx. 23 min with a 4480 mAh lithium polymer battery. The full setup of the quadcopter including camera and battery is 1.3 kg.

Aerial pictures of the full stretch are taken at 50 m altitude and lower, to ensure sufficient ground resolution of the photos. Additionally, 6 colored frames (0.5 m * 0.5 m) were placed in areas with different grainsizes below and above the water surface. These frames will be captured on the photos and serve as reference areas, where different approaches of grain-size distribution are tested including pebble counts and sampling of the top layer after all aerial data is ascertained. Water depth at each underwater reference area is measured in addition.

Photosieving can give a good impression about the grain size distribution, especially when the dominating substrate is gravel. Different algorithms of image processing will be used in this study and compared to each other.

In addition, all taken photos are merged to a georeferenced aerial picture. UAV based data in a birds eye view has the advantage of a better perspective compared to mapping in the field. The generated orthomosaic unfortunately is not sufficient for photo sieving purposes, as the full file is too big to handle with current software and in addition a ground resolution of 2 cm per pixel is not sufficient when it comes to areas with smaller fractions. Nevertheless a detailed substrate class map can be generated out of this orthomosaic, referenced based on photo sieving of its raw photos.

Photo sieving with aerial pictures works satisfyingly good. Especially the DJI System generated good footage as the camera was stabilized by a 3-axis gimbal. As the sensor of this setup is not per-forming as good as the one of the Sony NEX6 cam-era a lower flying altitude especially in the areas of finer fractions is necessary to produce a good dataset for photo sieving.