The addition of artificial intelligence software to camera technology has introduced a host of useful features that have enhanced how pictures and videos are taken – and how they are edited.
And it’s being used by amateurs on phones through to studio-level professionals on high-powered computers. It’s being deployed in the commercial space, too – in sectors including security/surveillance and logistics, where video footage is used to improve operational effectiveness.
“The key focus of Cartrack’s AI Vision solutions is to empower companies to establish a safety-first culture in their fleets by preventing accidents and ensuring company policies are embraced,” Cartrack said in response to a query by TechCentral. “For example, our AI Vision solutions allow our customers to reduce distracted and fatigued driving, mobile phone usage, seatbelt violations and unauthorised vehicle occupancy.”
The AI models used in fleet management are trained to monitor driver behaviour by identifying signals such as “eye fluttering” – a signal of sleepiness – or abnormally long gazes that could indicate that a driver has “zoned out”. Facial expressions that show shock or distress, meanwhile, could mean the driver is in danger. The AI software reports these insights to the logistics manager or command centre.
On one hand, the use of AI has reduced operational overheads that logistics companies would have to invest into driver monitoring, which in the past included the use of spotters along a vehicle’s path that would report what they saw to a central command centre.
This may give the impression that the technology is muscling in on work previously done by humans and threatening jobs, but the way in which fleet management providers are using it counters this narrative.
AI, human collaboration
Netstar, an Altron Group subsidiary, uses a fleet management bureau in Pretoria to centralise its operations across all its clients. All alerts sent by the AI software monitoring its vehicles are sent to the bureau where a team of trained technicians monitors and assesses the alerts, deciding which ones might require additional action. This collaboration between AI and trained human specialists helps minimise negative outcomes in a crisis.
“By introducing two-way communication through Netstar’s cameras, the operator can engage with the driver without the driver having to take their hands off the steering wheel. The operator then instructs the driver on corrective behaviour before providing escalation to the client or assigned respondent,” said Mark Forbes, GM for fleet bureau services at Netstar.
Other sectors use the technology in similar ways, though the AI is trained to identify a different set of behavioural patterns specific to the operating environment.
Read: Netstar launches global fleet management bureau in Gauteng
Worker safety is an important aspect in factories, warehouses and data centres. Here, AI cameras monitor workers to track the proper use of safety equipment, help minimise stock theft and scan the external environment to identify loiterers who may pose a threat to the facility.
The integration of AI into a facility’s security system also allows for corrective action to be taken prior to any human intervention. For example, an intruder spotted in a specific part of a building can be locked in by preventing security doors from opening to let them out, while an alert is sent to security personnel.
Using AI in this way does present some downsides. Video footage is data intensive, requiring far more storage capacity than text-based information. The processing done by the AI adds an additional draw on computing resources. Both of these factors amount to additional costs to the business. If the AI processing is done in the cloud, latency can be an issue, too.
Huawei recently launched an automated logistics facility in Johannesburg where more than 200 AI-powered cameras are used to manage security. According to Alvin Korkie, principal business and strategy consultant for sub-Saharan Africa at Huawei, the company opted for cameras that have on-device AI processing to minimise data processing and storage costs while also reducing latency.
“The intelligence is housed inside the camera, whereas typical solutions have a backend system that does the heavy lifting,” said Korkie. “We are able to detect running, intrusion, approaches to the fences and if someone tries to scale the fence. We also have crowd-density control, so if the number of people gathered outside the facility reaches a set threshold, an alert is sent to security.”
Marcel Bruyns, head of sales for sub-Saharan Africa at Axis Communications, a network camera specialist, said edge processing has other advantages, including improved analytics. This is because AI processing is done on raw footage, when the image is at its highest quality, before it is compressed and sent to an on-premises server or to the cloud.
For fleet management services, AI cameras have had some disadvantages. According to Netstar’s Forbes, AI use has led to a surge in the number of incidents detected, increasing the number of calls between the bureau, its clients and their drivers – and ultimately inflating the cost of communication.
The rise in incident detection has also contributed a higher number of disciplinary actions against drivers as companies try to plug performance gaps.
On the upside, improved training has decreased the number of adverse road incidents, leading to better long-term outcomes for companies running fleet operations. Forbes said the short-term spike in costs is outweighed by longer-term improvements in driver safety, decreases in road accidents and more efficient delivery runs.
‘No real drawback’
Generative AI has the tendency to hallucinate, leading to instances where false positives are generated where no corrective actions are required. According to Cartrack, this is where the fleet manager’s interaction with the AI becomes even more important, as it allows for feedback, prompting the machine learning models to improve their accuracy and deliver better results in future.
Read: Discovery turns to AI for ‘hyper-personalised health care’
“Many of our customers prefer to receive all detected events so they can review and determine which require action. Since it’s quick and easy for fleet managers to dismiss irrelevant events, there is no real drawback,” said Cartrack. – © 2025 NewsCentral Media
Get breaking news from TechCentral on WhatsApp. Sign up here.
Don’t miss:
Beyond instinct: how AI is reshaping retail store layouts in South Africa
