VC’s Distinguished Awards

Home » Engineering & Built Environment » PROCESS ENERGY & ENVIRONMENTAL TECHNOLOGY STATION (UJ PEETS) » Features » VC’s Distinguished Awards

​Vice Chancellor´s Distinguished Awards 2020: Innovation of the Year

The UJ PEETS team of Samson Masebinu, Nickey Janse van Rensburg, Tafadzwa Makonese, Katharina Gihring and Pieter Erasmus wins the Vice Chancellor’s Distinguished Award Innovation of the Year for 2020.


Empowering Cities with Data – Application of Artificial Intelligence (AI) for Feature Extraction from Aerial Imagery was implemented for the City of Johannesburg to enable them, for the first time, to identify, locate, count and estimate the installed capacity of all solar photovoltaic and solar water heater systems installed across the City.


City managers have limited resources to investigate, monitor, collect, and respond to the almost daily changes of the city landscape. To investigate some of these changes, aerial imagery of the city is captured at regular intervals and places of interest are manually analysed. A data-driven decision was needed for their energy demand-side management to reduce the drudgery of analysing these changes manually. An AI model was developed and applied for the City of Johannesburg, extracting from the aerial imagery of the City of Johannesburg.

The City of Johannesburg and the Department of Housing installed about 76,000 Solar Water Heaters (SWH) for households between 2008 and 2016 within the City to reduce peak demand for electricity due to the stressed generation capacity of Eskom. Lack of records on the specific location of these SWHs has limited future rollouts and any performance assessment. City Power, the electricity utility company of the City of Johannesburg, reported that they have noticed a decline in electricity demand, but could not confidently attribute the reduction to their energy efficiency strategies, activities of residents of the city, or its operational losses. They had, however, noticed increased adoption of solar photovoltaic (SPV) installations across the City.

The National Energy Regulator of South Africa (NERSA) requires the City to provide a database of installed SPV. Unfortunately, the City did not have sufficient data to report to NERSA as households with SPV are reluctant to register their installation with City for several reasons, such as paying fees, monitoring and security risk. Furthermore, the City lacks the means to comb the length and breadth of its landscape to document and report all installed systems. Hence, the City needed an alternative low-cost tool of data acquisition on installed SWH and SPV within its jurisdiction.


Responding to the City of Johannesburg´s need for a database with minimal cost and considering COVID-19 restriction, an AI approach was proposed since the objects of interest are installed in places exposed to sunlight and captured in any aerial imagery. Fortunately the City Geographic Information System (GIS) department captures the aerial imagery of the City every 5 years. The AI approach was based on convolutional neural network (CNN) model which was developed to analyse the high-resolution imageries of the City to identify, count, and geo-locate all installed SWH and SPV systems within the City. The CNN architecture was based on the dynamic UNET structure with ResNet 34 model used with ImageNet weight initialization to train the CNN for segmenting by pixel, the image region belonging to SWH and SPV objects respectively (see Figure 1). Furthermore, the AI approach was used to further estimate the surface area of the installed SPV system to estimate the installed capacity.

Visual Representation Cnn
Figure 1Visual representation of the CNN from high resolution imageries to identifying the SWP and SPV systems.


The novel approach implemented in this project is of interest due to its potential rapid deployment to other municipalities. A website-based model is planned where municipalities can upload their aerial imagery and with the click of a button intelligently identify all SWH and SPV systems installed within their jurisdiction. This information can be used to understand their Greenhouse emission footprint, the installed energy reduction measures and alternative energy potential.


The value propositions anticipated from this AI approach include:

  • Making well-informed decision based on aerial imagery;
  • Reduce the risk and cost associated with field work;
  • Optimize maintenance strategies for green infrastructure investment ;
  • Identify vacant rooftop spaces with the potential for SPV installation;
  • Identify illegal waste disposal sites; and
  • The redesign of commercial districts and urban planning (considering all flows of people, transport and goods) space can be optimally used.

We would like to acknowledge that this project is funded by C40 Cites and supported by the City of Johannesburg.

UJ PEETS is funded by the Technology Innovation Agency to enable technology innovation support for SMEs promoting circular green technology solutions.

High Resultion Uj Peets Logopng

Latest News

| View All News