Machine learning helps protect Africa’s wildlife

Read time 4min 40sec
Comments (0)

A partnership between German-based software giant SAP and non-profit organisation Elephants, Rhinos and People (ERP) is using machine learning to protect endangered species in Africa.

In the fight to protect Africa’s endangered wildlife, powerful tech such as cloud computing, drones and machine learning can play a critical role in enhancing conservation efforts, and in combatting poaching and the resultant illicit trade.

According to Rudi de Louw, head of the Co-Innovation Lab at SAP Africa, advances in technology and new ways of collaborating are paving the way for the development of innovative solutions to protect Africa’s wildlife.

“Having powerful technology means nothing if you can’t achieve accuracy and consistency in the data and outcomes it produces,” De Louw says.

“We undertook an extensive and ongoing co-innovation project with ERP focusing on technical feasibility, product development and refinement. The outcomes are exciting and potentially game-changing – especially, in this case, for an elephant and rhino population that remains under threat.”

Crisis levels

SAP says there are an estimated 30 000 rhinos remaining in the wild today, a significant decrease – largely due to poaching and habitat destruction – from the half million that roamed Africa and Asia at the start of the 20th century.

It says rhino poaching reached crisis levels in the last decade, with instances in SA increasing more than 9 000% from 13 in 2007 to 1 215 in 2014.

Encouragingly, the latest reported figures show a sharp decline in rhino poaching activity, with 594 cases in 2019.

Elephants have also been targeted, says SAP. After a single poaching incident in 2014 – the first in SA in a decade – elephant poaching activity spiked, with 71 reported cases in 2018.

The latest official figures show a decrease of more than half for 2019. Across Africa, thousands of elephants have been brutally slaughtered in recent years, with almost entire populations being wiped out in certain African countries, SAP notes.

ERP was founded to preserve and protect Southern Africa’s wild elephants and rhinos through a strategy that is based on alleviating poverty in rural areas surrounding the threatened creatures.

It forms part of the structure of, a largely employee-owned group of companies, non-profits and impact investment organisations with a strong global presence that also includes EPI-USE, an independent SAP HCM and payroll specialist.

Since 2017, ERP has piloted an anti-poaching strategy that has completely eliminated poaching of megafauna in the areas it monitors.

David Allen, ERP Air Force project lead at ERP, and a senior SAP practitioner in EPI-USE, says the deployment of new technologies has been a core element of the initiative's success.

“Following a year-long testing and innovation process in partnership with the Co-Innovation Lab, we have made major strides in refining our machine vision, machine learning and response capabilities.”

An unmanned aerial vehicle (UAV) was initially used to monitor the movement of elephants but was quickly deployed to provide a layer of intelligence to how teams responded to alerts.

Allen says one of the first priorities for the project was to address occasional instabilities in their prototype IT environment.

“The Co-Innovation Lab team helped us migrate to SAP Cloud Platform, and SAP is providing three years’ cloud hosting to support the production process.”

Siddharth Taparia, SVP and head of experience marketing at SAP, who led the team, says the project marked a world-first for the organisation.

“While we have supported non-profit organisations by providing our on-premises solutions before, this was the first time we supported a partner in the cloud and illustrates how the brave new world of cloud is transforming businesses of all sizes.”

Following the migration, ERP and the Co-Innovation Lab started working on technical feasibility tests for some of their more ambitious ideas, many of which have over time proven invaluable to the success of conservation efforts.

Improving response time

One of these ambitious ideas, according to De Louw, involves extensive development of machine learning algorithms to enable the team to improve its response, data capturing and processing capabilities.

“A network of different cameras within the reserve trigger whenever movement is detected. Machine vision is used to track movement while machine learning algorithms help distinguish between threats and non-threats. This has required us to feed our algorithms vast amounts of data to train them to distinguish between animals, people and other movement.

This accuracy is important; as soon as rangers receive too many alerts – especially if they prove to be false – trust in the system starts eroding, Allen says.

“We need to reduce false-positives to ensure rangers are only alerted when something requires their attention. This is simpler to accomplish with the on-the-ground cameras, but our UAV-mounted cameras require significant further training and development.

“A rhino seen from ground level is fairly easy to distinguish from an elephant or person, but as soon as you take an aerial view, animals tend to fade into the landscape. As our data set grows and we refine our machine learning algorithm, these types of inaccuracies will be resolved over time.”

Login with
17 Aug
Be the first to comment
See also