A labor-free index-guided semantic segmentation approach for urban vegetation mapping from high-resolution true color imagery

Accurate and timely information on urban vegetation (UV) can be used as an important indicator to estimate the health of cities. Due low cost RGB cameras, true color imagery (TCI) has been widely for high spatial resolution UV mapping. However, current index-based classifier-based mapping approaches face problems poor ability accurately distinguish reliance massive annotated samples, respectively. To address this issue, index-guided semantic segmentation (IGSS) framework is proposed in paper. Firstly, a novel cross-scale index (CSVI) calculated by combination TCI Sentinel-2 images, value provide initial map. Secondly, reliable non-UV samples are automatically generated training model, then refined map produced. The experimental results show that CSVI outperformed existing five indices highlighting cover suppressing complex backgrounds, IGSS workflow achieved satisfactory with OA 87.72% ∼ 88.16% F1 score 87.73% 88.37%, which comparable fully-supervised method.