This study utilizes advanced sensing techniques and artificial intelligence to enhance the efficiency and effectiveness of tree risk assessment in Hong Kong. Collecting tree information for risk assessment is a time-consuming task for arborists. To alleviate their workload, the first part of this study introduces an automated tree inventory solution that employs artificial intelligence algorithms to segment individual trees from point cloud and image data obtained through a mobile mapping system. Additionally, a novel method for estimating the diameter at breast height of trees with irregular shapes is proposed. While visual tree assessment is widely used for tree risk assessment, it captures only the tree's condition at a particular moment, potentially missing ongoing changes or future...[ Read more ]

This study utilizes advanced sensing techniques and artificial intelligence to enhance the efficiency and effectiveness of tree risk assessment in Hong Kong. Collecting tree information for risk assessment is a time-consuming task for arborists. To alleviate their workload, the first part of this study introduces an automated tree inventory solution that employs artificial intelligence algorithms to segment individual trees from point cloud and image data obtained through a mobile mapping system. Additionally, a novel method for estimating the diameter at breast height of trees with irregular shapes is proposed. While visual tree assessment is widely used for tree risk assessment, it captures only the tree's condition at a particular moment, potentially missing ongoing changes or future issues. In the second part of this study, an Internet of Tree Things system is adopted to enable long-term monitoring of tree stability. The collected tree tilt data is input into a deep learning model for anomaly detection. A four-level likelihood rating of tree failure, based on the number of AI-detected abnormalities within a time window, is introduced for risk analysis. Long-term stability trends are examined by comparing tree tilt angles under similar weather conditions. Moreover, the dynamic properties of trees have the potential to provide valuable insights into their structural integrity. However, the wind-tree interaction is not fully understood due to limitations in the spatial and temporal resolution of current measurement technologies. Therefore, the last part of this study incorporates a multi-beam flash LiDAR sensor to explore tree dynamics properties. This sensor accurately captures the motion of the tree at every position, enabling the derivation of dynamic properties which are validated through data collected from pull-and-release tests and during typhoons. In summary, this study presents an innovative approach to complement tree risk assessment to ultimately ensure the safety of urban environments.