# Waste Detection and Change Analysis based on Multispectral Satellite Imagery

**Authors:** Dávid Magyar, Máté Cserép, Zoltán Vincellér, Attila D. Molnár  
**arXiv:** [2303.14521](https://arxiv.org/abs/2303.14521)  
**Date:** 2023-03-25  
**Summary page:** https://www.summarizepaper.com/en/paper/2303.14521v1/

## Key points

- Illegal landfills and waste in rivers are major environmental issues
- Satellite imagery and machine learning can be used to detect and monitor these waste sites
- Two forms of waste detection: identifying hot-spots (illegal waste dumps) and identifying water-surface river blockages
- Medium to high-resolution multispectral satellite imagery is used, with a focus on the Tisza river as the study area
- Advancements in remote sensing and Earth observation satellites equipped with multispectral sensors are discussed
- Specific satellite systems used in the study: Sentinel-2 and PlanetScope
- Sentinel-2 has a high-resolution multispectral instrument with 13 spectral bands, including bands for vegetation analysis
- PlanetScope is a constellation of satellites capable of recording almost the entire land area of Earth daily
- Bands with resolutions of 10 and 20 meters from Sentinel-2 were primarily used for identifying large landfills
- Traditional methods of identifying illegal waste deposits require manual surveys, which are not scalable due to reliance on human labor resources
- Proposal to utilize satellite imagery combined with machine learning techniques for effective detection and monitoring
- of illegal landfills and waste in rivers
- Results will be presented in Section 4 followed by a summary of findings and conclusions in Section 5.

## Summary

The paper focuses on the problem of illegal landfills and waste in rivers, which are major environmental issues. The authors propose using satellite imagery and machine learning to detect and monitor these waste sites. They specifically analyze two forms of waste detection: identifying hot-spots (illegal waste dumps) and identifying water-surface river blockages. The study uses medium to high-resolution multispectral satellite imagery, with a focus on the Tisza river as the study area. In the background and related work section, the authors discuss the advancements in remote sensing and Earth observation satellites equipped with multispectral sensors. They highlight two specific satellite systems used in their study: Sentinel-2 and PlanetScope. Sentinel-2 is part of the European Union's Copernicus program and consists of two identical satellites that continuously monitor changes on Earth's surface. It has a high-resolution multispectral instrument with 13 spectral bands, including bands in the "red edge" range for vegetation analysis. PlanetScope is a constellation of satellites from Planet Labs Inc., capable of recording almost the entire land area of Earth daily. It comprises more than 130 satellites that provide high-resolution imagery. The authors explain that they primarily used bands with resolutions of 10 and 20 meters from Sentinel-2, suitable for identifying large landfills. They also mention that traditional methods of identifying illegal waste deposits require manual surveys, which are not scalable due to their reliance on human labor resources. To address this issue, this paper proposes utilizing satellite imagery combined with machine learning techniques for effective detection and monitoring of illegal landfills and waste in rivers. The results obtained from their study will be presented in Section 4 followed by a summary of findings and conclusions in Section 5.