Main menu
Numerous vehicles travel on Swiss motorways every day. These main traffic routes are used not only by people but also by alien plant species, which can spread along these corridors. Some of these so-called invasive neophytes reproduce rapidly, and may be toxic, cause allergies or lead to increased road maintenance costs. It is often unclear where and how quickly they spread as mapping them on motorways is dangerous work, which why such maps are lacking.
The WSL ecologist, Michael Nobis, and his team are now testing, in collaboration with the Computer Vision Lab at ETH Zurich, a new method to map plant species quickly and automatically. It involves researchers driving along a motorway, such as the A1 between Geneva and St. Margrethen, with two cameras to film the side and central strips.
Travelling at 90 km/h and recording 24 frames per second resulted in a data set with several million individual images of the vegetation along the roads.
This data set is then evaluated with the help of ‘Deep Learning’, i.e. with artificial neural networks that are trained to recognize certain patterns in data. On a comparatively small number of the pictures, the occurrence of tree-of-heaven and narrow-leaved ragwort plants are recorded by hand. With this training data set, the computer learns to identify the species. Michael is convinced: “the new technology could simplify time-consuming routine tasks such as mapping in the field.” An earlier project involving drones on SBB railway lines has shown that the machine sometimes identified the species better than a botanist.
The aim of the current project is to test the new method and produce detailed distribution maps of the tree of heaven and the narrow-leaved ragwort along motorways.
The Federal Roads Office (FEDRO), which commissioned the project, and the Federal Office for the Environment (FOEN) are very interested in the results. These will show the current distribution of invasive species along Swiss motorways and provide help in deciding how best to deal with these species.
(Lisa Bose, Diagonal 2/19)