OPERATIONAL RELIABILITY OF UAV-BASED WASTE SITE DETECTION FOR ENVIRONMENTAL MONITORING
FIABILITATEA OPERAȚIONALĂ A IDENTIFICĂRII ZONELOR CU DEȘEURI UTILIZÂND UAV-URI ÎN MONITORIZAREA MEDIULUI
DOI : https://doi.org/10.35633/inmateh-79-82
Authors
Abstract
Illegal dumping in peri-urban and natural environments poses persistent challenges for environmental authorities, particularly when large areas must be inspected with limited personnel and budget. Reliable and scalable monitoring approaches are essential for practical environmental management. Unmanned Aerial Vehicles (UAVs) offer rapid spatial coverage, yet it remains unclear how complex the analysis must be to reliably detect waste sites in practice. This study evaluates the operational reliability of three UAV-based image analysis approaches – global statistical screening, classical change detection with color filtering, and deep learning object recognition – under controlled field conditions. The field experiment used controlled contamination conditions so that the evaluation reflects practical detection reliability rather than algorithm benchmarking. The statistical method could not reliably identify localized waste, while change detection achieves moderate reliability but requires reference imagery. The deep learning model achieved the highest site detection reliability (~93%) and worked without baseline images. Its main advantage was improved object completeness rather than better identification of contaminated areas. These observations show that monitoring workflow design matters more than increasing algorithmic complexity. A tiered strategy that combines broad screening with focused high-accuracy analysis can reduce inspection effort without compromising the results of inspection. The framework helps to select monitoring approaches according to available resources, supporting scalable and cost-effective environmental monitoring programs.
Abstract in Romanian



