In the quest for fewer traffic jams and improved air quality in Taipei, the public sector has been exploring various traffic light management solutions for a number of years on the way to becoming a smart city. The congested city has over 2,500 intersections with traffic lights and, during rush hour commuters, needlessly spend hours in traffic. To address this, the transportation bureau commissioned a smart transportation solution provider to develop a traffic light surveillance system that would help improve traffic flow. To optimize their existing system further, the company leveraged ADLINK’s latest artificial intelligence (AI) enabled machine vision system, the EOS-JNX-I.
Currently, traffic flow information is picked up by the Power over Ethernet (PoE) cameras mounted on the traffic lights at each intersection and transmitted to a central control room for management. But during rush hour, there are not just additional cars to consider – there is also a significant uplift in the number of pedestrians wanting to cross the road. At some of the busier intersections, to accommodate the safety needs of both the cars and pedestrians with the respective green lights, this often results in longer red lights, slowing down the traffic and pedestrian flow considerably. There are also occasions when there are no pedestrians, but drivers may see a red light and have to wait unnecessarily at the intersection.
The smart transportation solution provider, therefore, wanted to enable AI on the existing surveillance system to make it more flexible. By using AI to analyze the traffic conditions, the operator can control the traffic lights automatically, according to the traffic flow and the number of pedestrians.
Enabling AI in an existing surveillance system may need an AI computer and an extra PoE hub, causing a higher total cost of ownership (TCO), as shown in Figure 1. In addition, there were concerns about losing the video stream from the IP camera to the network video recorder (NVR), if, for instance, the system crashes or reboots for some unexpected reason.
With its embedded uplink port, ADLINK’s EOS-JNX-I edge AI vision system enables edge AI safety capabilities with minimal change, cabling and investment. Moreover, with 100m cable validation, the EOS-JNX-I eliminates compatibility and reliability issues, so AI developers can easily apply AI inferencing with their image source.
Thanks to the open network video interface forum (onvif®) protocol, the EOS-JNX-I can connect any IP camera and any NVR, even if they are from different manufacturers. The system integrator selects the onvif option through the channel settings to discover each IP camera on the surveillance network. Switching the chosen IP camera to the onvif protocol enables the connection.
One of the biggest challenges to overcome is to ensure there is no interruption to the video stream. This can happen for two reasons – the camera may have crashed because it has overheated due to hot weather conditions. The other possibility for camera failure is that the cable has broken or come loose. When this happens the surveillance system provider needs to send maintenance personnel to the traffic light to check whether the camera has crashed or not, or to check the cable status, which costs time and money. With the EOS-JNX-I, ADLINK provides the smart PoE to solve these two problems.
The ‘golden rule’ of IP camera troubleshooting is to reboot it. Also, sometimes operators need to manually turn on/off the camera to configure some software settings. The smart PoE capability enables operators to reboot the camera remotely, minimizing the risk of manually plugging/unplugging the power cable and effects of downtime. If the camera does not reset after power on, operators may judge the issue is caused by another issue and prepare a different tool for maintenance.
ADLINK also offers PoE loss detection, which provides alerts if the power is unexpectedly cut off, enabling swift repair and maximizing uptime.
Through statistical analysis, the intersections of the upgraded system did reduce unnecessary red lights and the degree of congestion. Over time, with more video live stream data, the machine learning (ML) models will mature to identify more patterns, enabling the system to make more accurate predictions and ease congestion further.
ADLINK’s EOS-JNX-I AI-enabled vision system provides easy maintenance, ultra reliability, and fast development, makes it an ideal platform for any IP camera surveillance system in application scenarios far and wide, including safe communities, industrial inspection, smart retail, factory logistics, and delivery robots.
For more information, please visit ADLINK EOS-JNX PoE AI-Vision System page.