Implementing Deep Learning algorithms for urban tree detection and geolocation with high-resolution aerial, satellite, and ground-level images

• DOI: 10.1016/j.compenvurbsys.2023.102025
• Publication Date: 2023-08-14T15:43:57Z
• Authors: Luisa Velasquez-C...; Maddi Etxegarai; Sergio de-Miguel
• Keywords: Geolocation; Tree (set theory); Urban forest; Aerial imagery

Urban forests are becoming increasingly important for human well-being as they provide ecosystem services that contribute to improving of city dwellers and addressing climate change. However, despite their importance, there is an information gap in most the world's urban due high cost complexity conducting standard forest inventories environments. New technologies based on artificial intelligence can represent a smart efficient alternative costly traditional inventories. In this paper, we present approach deep learning algorithms detection, counting, geopositioning trees using combination ground-level aerial/satellite imagery. We tested several convolutional networks, exploring different combinations hyperparameters adjusting query distance between images, detection radius, various resolutions satellite aerial images. Our methodology able detect accurately locate 79% street tree with positional accuracy 60 cm center canopy. Additionally, allows us determine availability photographs trees, indicating from which Google Street View image each visible. research provides scalable replicable solution scarcity data worldwide, demonstrating potential revolutionize way inventory monitor forests.