Development of automatic methods for the detection and differentiation of illegal waste dumps and plastic waste in the environment using hyperspectral satellite imagery and deep learning

The aim of HyPla is to develop free software tools - based on free satellite data - for mapping illegal waste dumps, including plastic waste accumulations, as part of the EnMAP-Box (Environmental Mapping and Analysis Program) software for use in science, education, administration and business. In particular, the proposed project's first use of hyperspectral satellite data for material characterisation on suspected waste sites in combination with an upstream, CNN-based (Convolutional Neural Networks), area-wide search for suspected sites based on the spatially higher-resolution channels of the S-2 sensor (synergistic use), taking into account the spatial context, could represent a breakthrough in this innovative field of application towards complete automation and the achievement of classification accuracies > 90%.