Elizabeth Moselakgomo left her house in Rhodesfield, Kempton Park, on the afternoon of Wednesday, 9 September, to go for a run. Her eight-year-old daughter was at home, waiting. When her mother did not come back, it was that child who phoned the relatives, and the relatives who went to the police.
Officers searched and made inquiries at hospitals and mortuaries and turned up nothing. On Saturday, 12 September, her body was found behind a hotel in the suburb, a short walk from her front door. Her family identified her at the Germiston mortuary on the Monday. She was 38. The people who loved her called her Tsontso.
It took three days to find her. Rhodesfield is not a remote place. It is minutes from OR Tambo, in the province that has more cameras trained on it than any other in South Africa.
Lessons from Asia
In April, the Seoul Metropolitan Government in South Korea commenced the roll-out of an artificial intelligence video search system across all 25 of its districts. It can search in the order of 123,000 cameras at once and work through 100 hours of footage in about 10 minutes. The reason the city gave for installing the system was to shorten the search for missing persons.
The same city runs a safe-return-home app tied to its municipal cameras, so that a woman walking home at night can be watched until she is inside and can get a control room operator to focus on her by shaking her phone.
India took a different route after the Delhi gang-rape case in 2012 and built what its Safe Cities programme calls safety islands, which are lit points along a route fitted with a camera and a button wired to a response vehicle.
I’m uneasy about using a dead woman to argue about procurement, and I want to say so plainly before I do it. But decisions like these get taken in the weeks when people are still angry, and then not again for years. So let me be specific about what I want bought, what it costs and whose signature it needs.
The cost
Start with what we have already paid for. In a written reply to the Gauteng Legislature earlier this year, the provincial government confirmed that R124.5-million had been spent on installing 960 CCTV cameras, and that 89 sites were non-functional, affecting 269 cameras. While the DA claimed this meant about 28% of the network was offline, the department dismissed the characterisation as misleading, maintaining that repair teams were actively fixing the failed sites. I'll leave them to that argument.
What neither side disputes is the reasons the cameras went down: vandalism, water damage to batteries and faulty inverters. A good share of the network failed on the same battery-and-inverter kit the rest of us bought to survive load shedding, and nobody costed in the replacement cycle.
The province puts ongoing maintenance at R12-million to R15-million a year. Pay it and publish an uptime figure so that residents can see whether the camera at the end of their street is watching anything at all. This is the cheapest item on my list, and the figures have been public since May.
Monitor routes
Then install cameras where women walk. The province says each camera costs about R130,000, so a first tranche of 300 along identified walking and running routes, station precincts and taxi ranks in Ekurhuleni comes to roughly R39-million.
Against the R1.9-billion the provincial e-government department holds for digital transformation, or the R6.8-billion Cape Town has put into safety this year, the order of magnitude is small. Anyone who drives across this province already knows what a well-funded camera network looks like. It looks like poles every couple of hundred metres in the suburbs that pay for them, and nothing much around a station precinct at 5.30am. Where the new cameras go should be settled by a safety audit walked with the women who use those routes.
Link e-panic buttons to the cameras. Gauteng has distributed more than 184,000 e-panic buttons under its e-policing programme. Press one now, and a responder is dispatched to you. In Seoul, that same press also puts an operator onto the camera already pointed at where you’re standing, which is the part we have skipped.
We have built the expensive half of the system twice over, once in public cameras and once in private ones, and left the two halves unconnected. Joining the e-panic platform to the provincial network and the district control rooms is a software and integration job. The province should cost it, tender it and attach a date.
Expedite the search
This speaks to those three days. Seoul did not build new cameras to get its search capability. It bought analysis that runs over cameras it already had. Gauteng has its 960, and Vumacam runs several thousand more across Johannesburg, Tshwane and Ekurhuleni with analytics and licence plate recognition, to which the provincial government already holds an access agreement.
What we do not have is a written protocol that starts a coordinated video search the moment an adult woman is reported missing, with no informal waiting period, and that obliges the private networks to release relevant footage in hours rather than weeks. Writing that protocol costs nothing. It has not been written.
Four bureaucratic entities are involved:
The Gauteng Department of e-Government: Manages the e-policing programme, including maintenance and panic-button integration.
The City of Ekurhuleni Department of Community Safety: Holds authority over the pedestrian camera network.
The SAPS provincial commissioner: Authorises and controls the search protocols.
The Gauteng Department of Community Safety: Brokers terms and data-sharing agreements with private security networks.
This operational split requires no new legislation, nor does it require any one department to wait for another to act.
Objections
I should note the objections, because I’ve heard all of them and some of them are right.
Cameras record; they do not intervene. Footage is worth its cost only when somebody is watching the feed and another person can get there in minutes, which is why control room staffing and vehicles have to be funded alongside the hardware, and why the state of the network we own matters more than the next procurement announcement.
Facial recognition misidentifies people, and in a country with our history, the power to track a face across a city needs a narrowly defined purpose, external oversight and strict compliance with our own data protection law. Researchers looking at the Indian Safe Cities projects have warned that these systems drift into policing how women move while claiming to protect them. I don’t dismiss that. I’ve watched technology sold in this sector as a substitute for capability more than once.
Multiple murders
By Thursday, 17 September, the bodies of nine murdered women had been found in and around Kempton Park since 15 July, and the linkage work that might have connected one docket to another only got going after the fifth body was discovered and residents took to the street.
What a camera network would have done is shorten the three days it took to discover Elizabeth Moselakgomo’s body. Somewhere there is almost certainly footage of her on that Wednesday afternoon, jogging out of her own street. Nobody could get to it in time. DM
Comments
Scroll down to load comments...