Greenery Scanner: A drive-by platform for urban greenery NDVI sensing using low-cost multispectral camera

Urban trees provide environmental, social, and economic benefits, but they are increasingly affected by biotic and abiotic stressors. Frequently updated tree-health inventories are therefore essential for proactive urban greenery management. Aerial and satellite multispectral imaging can support tree-health assessment through the Normalized Difference Vegetation Index (NDVI), but in dense urban environments these approaches may be limited by canopy occlusions, top-down viewing perspective, operational costs, and revisit constraints. Street-level multispectral imaging offers a complementary perspective by capturing lower and lateral canopy portions that may contain stress signals invisible from the above. This study proposes a method for street-level urban greenery monitoring, enabled by Greenery Scanner (GS): an open-source, low-cost, drive-by platform. The approach combines vehicle-mounted multispectral sensing nodes with an AI-based processing pipeline that segments tree canopy pixels from panoramic images and extracts calibrated NDVI values. We use the platform for crown dieback detection in Pinus pinea trees in Rome, Italy, where canopy decline is primarily associated with infestation by Toumeyella Parvicornis . A laboratory experiment measured the functional and radiometric consistency of GS against a Spectralon target with a mean NIR/R ratio of 1.037 and an NDVI residual error of 0.018 under controlled illumination. A field deployment on 84 trees showed that GS-derived NDVI was significantly associated with arborist-assessed crown dieback severity (p<0.01). The data was used to feed a binary logistic classifier distinguishing severe (≥50%) from early/low dieback (<50%) achieving 66% accuracy. We discuss how GS can support prioritizing arborist inspections, complementing and enhancing existing methods.