A considerable amount of the plastics produced around the world is now dispersed throughout the environment, and in particular in aquatic ecosystems. This can have damaging consequences for plants, animals and human beings. This study investigates some approaches for detection and monitoring of plastics waste in river habitats through multispectral image classification. The data are acquired using a proximity sensor in the electromagnetic spectrum range that includes the ultraviolet, visible and near infrared bands, as for the WorldView‐2 satellite. The in‐depth analysis of the spectral signatures obtained shows typical plastics trends and reflectance values in the near infrared bands. Different classification methods were compared to test their effectiveness for the isolation of plastics samples dispersed in a river habitat. This project represents the first step within a wider research programme, with the aim to define a new approach for future river pollution monitoring.

Abstract. Plastic is the third world’s most produced material by industry (after concrete and steel), but people recycle only 9% of plastic that they have used. The other parts are either burned or accumulated in landfills and in the environment, the latter being the cause of many serious consequences, in particular when considering a long-term scenario. A significant part the plastic waste is dispersed in the aquatic environment, having a dramatic impact on the aquatic flora and fauna. This motivated several works aiming at the development of methodologies and automatic or semi-automatic tools for the plastic pollution detection, in order to enable and facilitate its recovery. This paper deals with the problem of plastic waste automatic detection in the fluvial and aquatic environment. The goal is that of exploiting the well-recognized potential of machine learning tools in object detection applications. A machine learning tool, based on random forest classifiers, has been developed to properly detect plastic objects in multi-spectral imagery collected by an unmanned aerial vehicle (UAV). In the developed approach, the outcome is determined by the combination of two random forest classifiers and of an area-based selection criterion. The approach is tested on 154 images collected by a multi-spectral proximity sensor, namely the MAIA-S2 camera, in a fluvial environment, on the Arno river (Italy), where an artificial controlled scenario was created by introducing plastic samples anchored to the ground. The obtained results are quite satisfactory in terms of object detection accuracy and recall (both higher than 98%), while presenting a remarkably lower performance in terms of precision and quality. The overall performance appears also to be dependent on the UAV flight altitude, being worse at higher altitudes, as expected.

Abstract. Plastic pollution has become one of the main global environmental emergencies. A considerable part of used plastics materials is dispersed or accumulated in the environment with a significant damaging impact on many terrestrial and aquatic ecosystems.Artificial Intelligence has proven a fundamental approach in last years for the detection of plastics waste in the aquatic habitats: several groups have recently tried to tackle such problem by developing some machine learning-based methods and multispectral or RGB imagery. This study compares the results obtained by two machine learning classifiers, namely Random Forests and Support Vector Machine, to detect macroplastic in the fluvial habitat through multispectral imagery. The acquisition of images has been made with a hand-held multispectral camera called MAIA-WV2. Despite the obtained results are quite good in terms of accuracy in a random validation dataset, some issues, mostly related to the presence of white rocks and glares on water have still to be properly solved.

This paper shows the results of applying high-resolution Unmanned Aircraft System (UAS) photogrammetric surveying on a large landslide. A real case study, where permanently installing GCPs could be complex, where natural shaped and formed land pose severe limitations in deploying ground targets with optimal geometric configuration. We analysed performances in terms of survey accuracy obtained by performing photogrammetric surveys through the Zenmuse P1 DJI optical camera and Phantom 4 Pro 2. In combination with DJI Matrice 300 UAS, the P1 camera allows direct georeferencing through GNSS observations in RTK mode. Photogrammetric surveys, performed through different georeferencing methods, have been compared. Several targets have been permanently installed on the ground over the maximum vegetation height to guarantee long-lasting reference over the years in the area, which is characterised by a diffuse short vegetation coverage. Multitemporal UAS surveys have been then compared using Digital Image Correlation (DIC) algorithms, and deformation maps have been produced. Afterwards, DIC results were compared with observations made by the GNSS ground-based permanent receivers resulting in a standard deviation of 0,077 m. Through results analysis, good accordance between ground-based GNSS observations and DIC analysis on the photogrammetric surveys have been identified for the same time span. To conclude, this type of landslide presents a moderate deformation speed; in such a case, effective deformations monitoring could be achieved using pseudo-direct georeferencing, which permitted a 0.24 m accuracy on the whole tested area.

Abstract. Most of the anthropic pollution arriving to seas and oceans is carried by rivers, leading to a dramatic impact on the aquatic flora and fauna. Hence, several strategies have already been considered to reduce the use of plastic (and non biodegradable) objects, fostering recycling, detect litter in the environment, and catch it. This work aims at implementing a litter detection approach usable also in urban areas, hence considering a mini-UAV (Unmanned Aerial Vehicle) in order to reduce the issues related to flight restrictions. The implemented strategy exploits a high resolution map of the area of interest, a properly trained deep learning litter object detector, and a vision based localization system, which allows to remarkably reduce the positioning error of the UAV navigation system, in order to provide estimates of the litter object positions with an accuracy at decimeter level for objects not too far from locations recognizable in the map.