UBC Theses and Dissertations
Mapping urban trees with deep learning and street-level imagery Lumnitz, Stefanie
Planning and managing urban trees and forests for livable cities remains an outstanding challenge worldwide owing to scarce information on their spatial distribution, structure and composition. Sources of tree inventory remain limited due to a lack of detailed and consistent inventory assessments. In practice, most municipalities still perform labor-intensive field surveys to collect and update tree inventories. This thesis examines the potential of deep learning to automatically assess urban tree location and species distribution from street-level photographs. A robust and affordable method for detecting, locating, classifying and ultimately, creating detailed tree inventories in any urban region where sufficient street-level imagery is readily available was developed. The developed method is novel in that a Mask Regional Convolutional Neural Network is used to detect and locate tree instances from street-level imagery, creating shape masks around unique fuzzy urban objects like trees. The novelty of this method is enhanced by using monocular depth estimation and triangulation to estimate precise tree location, relying only on photographs and images taken from the street. In combination with Google Street View, a technique for the rapid de- velopment of an extensive tree genera training dataset was presented based on the method of tree detection and location. This tree genera dataset was used to train a Convolutional Neural Network (CNN) for tree genera classification. Experiments across four cities show that the novel method for tree detection and location can be transferable to different image sources and urban ecosystems. Over 70% of trees recorded in a ground-truth campaign (2019) were detected and could be located with a mean error in the absolute position ranging from 4m to 6m, comparable to GPS accuracy used for geolocation in classical manual urban tree inventory campaigns. The trained CNN classifies 41 fine-grained tree genera classes with 83% accuracy. The detection and classification models were then used to generate maps of urban tree genera distribution in the Metro Vancouver region. Results of this research show that developed methods can be applied across different regions and cities and that deep learning and street-level imagery show promise to inform smart urban forest management, including bio-surveillance campaign planning.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International