The objective of the project is to establish a tree management system based on computer modelling and AI to predict and identify trees posing high risk to railway assets, based on a number of attributes e.g. tree height, tree type, predicted growth rate, soil condition etc. Since the trees and vegetation are widely spread in different parts of the railway alignment with different terrain characteristics, the regular and consistent acquisition of tree data as well as the choice of attributes in the computer model is crucial to the success of prediction.

As any fallen tree or overgrown vegetation may lead to train service interruption, the system should be able to generate warning to the users well in advance before an undesirable event occurs, so that users can take advance measures to prevent such occurrence.

The project shall include:

1. Establish an AI-based computer model for predicting and identifying trees with high risk along the railway alignment
2. Collect tree data using satellite imagery and/or trainborne Light Detection and Ranging (LiDAR) sensors
3. Automatic pre-process the data for feeding into the AI-based computer model.
4. Analyse all collected historical and new data to identify and prioritize trees imposing risk
5. Adopts a feedback mechanism to improve accuracy of prediction
6. To make comparison between this new tree management system and traditional methods in terms of the effectiveness, efficiency, frequencies, cost, prediction accuracy of tree risk

12 months

-       1.5 months to study and design an AI-based computer prediction model,  together with identifying the gap against the tree data available from MTR
-       3 months for procurement of data acquisition system and software development of the prediction model
-       1 month for on-site installation and system testing
-       6 months for monitoring of prediction accuracy and refinement of modelling
-       0.5 month for evaluation and wrap-up