Urban Growth Monitoring in 3D : an innovative approach based on Lidar processing with Machine Learning

Modeling and monitoring urban growth in 3D are an innovative technology approach for planning systems and urban planning. The technology used for modeling is the Airborne Lidar, the reason is that, this smart tool provides « chronological data » on big cities such as Metropolitan sectors. Though, the solutions available today do not meet urban planning needs such as monitoring urban growth at the building level with a third dimension. To meet the challenges, two solutions are possible, which are developed in integrated procedure as Shirowzhan and Trinder tried to explain in their article :

  • Solution 1: Using a classification of buildings of an area by using Lidar time data « chronological data » to determine extent of the modifications comparing the heights of the buildings extracted. In this case, authors tasted two approaches: the first one is relied on a pixel-based approach based on a machine learning algorithm identified as a Support Vector Machines (SVM), the second one is based on a point-based approach, it operates with an ERDAS IMAGINE tool a software solution which allows a data process applying a remoting sensing and photogrammetry technologies.
  • Solution 2 : Using an algorithm for detecting modifications to lidar temporal datasets this make able to determine if modifications have taken place or not in a building classification. For this option, two pixel-based algorithms were used, specifically, SVM Support Vector Machines and image differentiation.

How does it work if a combination of the two solutions is possible?

Taking into account the advantage of SVM technology with the advantage of the image differentiation method to provide data and chronological data of the building high change’s magnitude. By considering an integration of the two solutions would solve few problems. For this approach, additional processing is needed after obtaining results. The detected changes must be set to which class they belong to for example; building class, vegetation class, road class, etc. Based on the purpose of the research or the study, and the urban planner’s aims it may be required to eliminate the items that are not essential to the analysis or the study. For example, if the 3D results of the buildings or the classifications contain major modifications it would be useful to remove all other classes from the integrated result.



With the integrated process (combination of the two solutions) result, urban planners would be able to determine changes with high buildings modifications instead of making a less precise estimations using the conventional techniques with a less precise results and details using solution 1 or solution 2.

This BIM and SIG combination give a precise vertical change allowing policy makers estimating the « mass on voids » and « buildings on green spaces » ratios according to the authors, which would enhance the use of airborne Lidar and may provide a big push to sustainable studies, analysis and researches.


Sources : 

Consultation of the sites on 12/01/2022, 12/05/2022 and 12/06/2022.

Sara Shirowzhan and John Trinder (2018), « Monitoring 3D Urban Growth: An Innovative Approach Integrating Lidar Processing with Machine Learning », Gim International [Online]

Sara Shirowzhan (2016), « Spatial and temporal pattern analysis of 3D urban development using airborne Lidar », PhD thesis, University of New South Wales, Sydney, Australia

Sara Shirowzhan and John Trinder (2017), « Building Classification from Lidar Data for Spatio-temporal Assessment of 3D Urban Developments », ScienceDirect [Online]

John Trinder and Mahmoud Salah (2011), « Support Vector Machines : Optimization and Validation for Land Cover Mapping Using Aerial Images and Lidar Data », Presented at 34th ISRSE Sydney, Australia, from 10-15 April 2011 [Online]