• Free Website Templates - Downlaod Full Themes
  • Topoequipos

    Soluciones integrales en GEOMÁTICA.

    Topoequipos Bogotá

    Calle 51 # 5 - 56 - Teléfonos: 338 12 26 - 288 02 03 - Esta dirección de correo electrónico está protegida contra spambots. Necesita activar JavaScript para visualizarla.

  • Equiposde Topografía

    Nuestros Productos y Servicios:

    Consultoría - Topografía -GPS - GNSS - Batimetría - Laboratorios didácticos - Fotogrametría - Software - Suelos - Drones - Accesorios - Franquicias - Estaciones Metereológicas - Equipos láser - Equipos para Construcción - Energías renovables

  • 1
  • 2

Monitoring 3D Urban Growth

An innovative approach integrating Lidar processing with machine learning

4be8874a75f75b611c8925d7c1a4f473f946d210Automated measurement of the third dimension of urban growth is a crucial requirement for future urban planning and monitoring. Airborne lidar is increasingly being used to produce regular time-based data over metropolitan areas. The current solutions ignore the specific needs of urban planners, are complicated to run and fail to address the problem of monitoring 3D urban growth over time at building scale. To address the current challenges, two possible solutions have been developed into an integrated tool.

Technical Challenges

Urban planners want to automatically estimate all three dimensions and model the building changes, their footprints and absolute height information over time to monitor the growth and predict the change patterns accurately. To monitor changes of buildings over time using airborne lidar data, there are two possible solutions to processing the data:

  • Solution 1: Classification of buildings using temporal lidar data sets and the determination of the changes of the extracted buildings from them. For this solution, two approaches were tested; a pixel-based approach using a machine learning algorithm known as Support Vector Machines (SVM), and a point based approach using an ERDAS tool.
  • Solution 2Application of a change detection algorithm on temporal lidar data sets and determining whether or not changes have occurred in a building class. For this solution, two pixel-based algorithms are applied, namely, SVM and Image Differencing.

This article tests which of these two possible solutions is preferred and which solution is able to best determine the changes in the magnitude of building heights.

Data Processing

To investigate the solutions at least bi-temporal airborne lidar data sets are required. In this research, the data sets used were collected in 2005 and 2008 over University of New South Wales (UNSW) campus, Sydney, Australia. Lidar data is usually represented by points; however, it also can be converted to pixels containing XYZ information of the points within the pixels. The resultant raster data is called a Digital Surface Model (DSM). The UNSW lidar dataset DSM covers plain, sloping and complex urban scenes with various sizes of urban objects (Figure 1).

Before applying any algorithm, some pre-processing is required. Removing outliers from the data sets is a key preparation task for both building classification (Solution 1) and change detection (Solution 2). In this case outliers can be generated by the measurement process (measurement noise) or constitute height points that do not contribute to the change detection process of urban buildings. Outliers observed in the 2008 data set include construction machinery such as tower cranes (Figure 2).

For Solution 1, SVM was applied to both the 2005 and 2008 datasets. In addition, the point-based classification tool in ERDAS software was used. For Solution 2, image differencing and SVM were used on the time series pixels.


The research showed that Solution 1 using building classification of time series lidar data for monitoring 3D urban growth is time consuming, cumbersome, complex for non-experts and less accurate than using a change detection algorithm. As demonstrated in figure 3, the building classification by SVM for 2005 data shows significant misclassifications between roads and buildings in sloping terrain, which is the result of classifying the boundaries of buildings as trees. This is in line with previous experimentation in the same area by the authors. This problem can probably be remedied by adding aerial image data and stacking the resulting RGB layers as additional layers for the SVM method. Since temporal aerial images were not available, further work on improving the SVM result using additional images was not possible.

Next, the classification tool in ERDAS was tested for its suitability for the point based approach for Solution 1 (building classification). The algorithm in ERDAS requires a number of parameters that have to be set interactively untill the best result is obtained for the area under study (table 1). With the chosen parameter thresholds, there was no problem of misclassification between roads and buildings in sloping terrain. However, visual comparison of classified buildings still showed some discrepancies among the results of unchanged building points (figure 4 and 5). Another inconsistency can be seen in the level of omission errors in the two datasets which is higher in 2005 than 2008. This inconsistency is an important issue for the determination of volumetric elements and pixel-based spatio-temporal building volume change calculation of an urban area over a time interval.



Min slope


Plane offset


Min height

0 m

Min area

100 m2

Max area

10,000 m2


0.3 m

Table 1: Parameters used in ERDAS

Solution 2 (change detection) using image differencing is less complicated than the procedure described above; however, with this method there are problems such as a high level of noise which causes a significant ‘salt and pepper effect’. For the SVM method applied to Solution 2, there is a lack of extraction of the magnitude of the height change. Building boundaries and missing data also cause errors that affect the results of DSM differencing.

Integrating Two Methods

Considering the advantage of SVM giving a smaller salt and pepper effect and the benefit of the image differencing method for providing the magnitude of height change, integrating them would address both issues, as demonstrated in Figure 6. For this solution, additional post-processing is recommended to determine that the detected changes belong to a specific class; building, vegetation, road, etc. Depending on the aim of the research, it might be necessary to remove all unwanted items. For example, if 3D changes of buildings are important for sustainability studies, all other classes should be removed from the integrated result.


With the post-processed integrated result, urban planners would be able to determine the changes in the magnitude of building heights rather than an overall less accurate estimation using either of the conventional techniques separately. Accurate vertical changes enable policymakers to estimate ‘mass-voids’ and ‘buildings to green space’ ratios, which consequently increase the application of airborne lidar for the built environment.


Shirowzhan, S., 2016, ‘Spatial and temporal pattern analysis of 3D urban development using airborne lidar’, PhD thesis, University of New South Wales, Sydney, Australia.

Shirowzhan, S. and J. Trinder, 2017. Building Classification from Lidar Data for Spatio-temporal Assessment of 3D Urban Developments. Procedia Engineering, DOI:10.1016/j.proeng.2017.04.308

Trinder J. C. and, M. Salah 2011 Support Vector Machines: Optimization and Validation for Land Cover Mapping Using Aerial Images and Lidar Data, Presented to 34th ISRSE Sydney, Australia 10-15 April 2011,
Miércoles 25 de Julio del 2018

Nuestros Servicios

  • Consultoría en ingeniería en General >

      Ofrecemos asesoramiento, desde una investigación preliminar, hasta la realización de un proyecto llave en mano, con la posterior colaboración en elmantenimiento y explotación, ejecutando todos los pasos intermedios. Los servicios que ofrecemos desde el nacimiento hasta la concreción de una ideason: • Identificación de proyectos • Estudios de viabilidad técnicos, socio-económicos y Read More
  • Servicio técnico Especializado >

    Servicio de mantenimiento y calibración de equipos topográficos Cuenta con personal especializado, con amplia experiencia enreparación, mantenimiento y verificación de equipos ópticos, mecánicos y electrónicos. Reparación de instrumentos de topografía y geomática (estaciones totales, niveles electrónicos, niveles digitales, niveles automáticos, gps, teodolitos, y otros) Contratos de mantenimiento preventivo y correctivo para Read More
  • Museo de la Topografía >

    Presentamos a ustedes una gran galería de imagenes e información de los diferentes equipos que han existido en este gran mundo de la topografía Read More
  • Mundo al instante >

    La finalidad de este espacio es presentarles a uds las ultimas noticias del sector,  y artículos de ingeniería que pueden ser de su interés Read More
  • 1
  • 2