Answers are grounded in this paper and cite their sources.
Demonstrates improved street tree detection and species mapping in urban areas, suggesting enhanced ecological management.
Accurate urban street tree recognition is crucial for ecological management. Street-view images offer a unique pedestrian perspective, capturing details often missed by remote sensing imagery. Current street tree recognition methods using street-view images are limited by scarce labeled data and high manual annotation costs. To address these challenges, we propose a novel urban street tree recognition and mapping method that enables accurate instance-level mapping of street trees. The method supports species identification where sufficient visual distinction exists. First, we design a two-stage street tree recognition method. In stage one, an object detection model is used to extract individual tree regions from street-view images, decoupling tree detection from species classification. In stage two, tree species are classified using a model pre-trained on open-source tree species datasets and fine-tuned with web-sourced labeled images, achieving annotation cost reduction. Second, StreetTreeMapper reduces duplicate detections by integrating street-view image features with OpenStreetMap’s geographic constraints, achieving precise tree instance mapping. To validate our proposed method, we construct a street tree dataset from New York, Zhuhai, and Seoul. Experimental results validate the overall effectiveness of our proposed method in both tree recognition and mapping. Through decoupled detection and classification, our method achieves a 2.7% to 11.4% increase in average precision compared to state-of-the-art models on Zhuhai datasets. The StreetTreeMapper achieves instance-level tree mapping with median errors of 4.5 m (New York), 4.8 m (Seoul), and 2.9 m (Zhuhai), reaching minimum observed localization errors of 0.21 m, 0.13 m, and less than 0.001 m, respectively. Our method enables automated detection and species identification of street trees. These capabilities provide valuable data support and decision-making foundations for smart city development and sustainable ecological maintenance.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: