Improving Incident Detection KPI on SANRAL’s Freeways in Gauteng

Abstract

The South African National Roads Agency Ltd (SANRAL) primarily relies on CCTV cameras to detect traffic incidents occurring on the Freeway Management System (FMS) network. On the Gauteng FMS network, over 90% of the incidents are detected using CCTV cameras. The operators have to manually pan, tilt and zoom each camera to detect incidents along the freeway. Traffic incidents are the major cause of severe or fatal injuries, congestion and delays on the freeway. They may also result in secondary incidents such as rear-end or multi-vehicle collision. It is therefore of utmost importance that the incidents are detected and cleared within the shortest time span. In the current contract, SANRAL has set the ‘Incident Detection KPI’ as 3 minutes. In other words, on average, incidents have to be detected within 3 minutes from the time of occurrence of the incident. Once the incident has been detected, the operator would rewind the video footage to determine the ‘occurrence time’ of the incident. However, in most cases (approximately 70%), the occurrence time of the incident is unknown. This is predominantly because the camera was facing away from the incident location (facing the opposite direction). This study aimed at improving CCTV surveillance, given the current infrastructure and resources; thereby increasing the number of incidents with an occurrence time. The study assumed that there would be no changes to the current camera positions, type of camera being used and operational structure. It was also assumed that there would be no additional cameras or human resources. Several surveillance methods were evaluated. The proposed surveillance method was tested using a before and after study. Incident data from May 2017 was used as the “before” and incident data from May 2018 (three months after implementation of the proposed new method) was used as the “after” period. The results of the analysis showed that subsequent to the implementation of the automated pre-set surveillance method, the number of incidents with an occurrence time increased by approximately 15% – an increase of approximately 500 incidents. The paper eludes to some of the shortcomings that still exist in the new method and possible ways of overcoming it.