In dense urban environments where traffic flow is heavy, keeping up with traffic control and maintaining road safety at intersections with the heaviest traffic are often tricky as vehicles and pedestrians’ behaviors are erratic and unpredictable, not to mention the fact that not all are prone to follow the traffic regulations at all times.
Traffic violations such as improper turns, speeding, median crossing and red-light running are daily occurrences, leading to even more serious offenses, worsening traffic congestions and increasing number of preventable accidents.
A city government came to Lanner looking for a hardware solution intended for curtailing the aforementioned violations; this particular hardware appliance was to be integrated with Gorilla IVAR™ and deployed as an AI-based intelligent video traffic monitoring system. Ultimately, the goal of this project does not end with preventing traffic violations, it must also deliver a blueprint for how similar solutions could be utilized for
- troubleshooting traffic issues found elsewhere in the country
- designing cities with better sustainability and public safety
- enabling data collection/analysis for large-scale infrastructure rollouts
An AI-based traffic monitoring solution involve the integration of rugged edge AI-ready platforms, AI inference servers and intelligent video analytics recorders. The said AI-ready appliance must be capable of, or enable, not only recording and capturing real-time traffic data such as vehicle speed and driver behavior, it would also be expected to pre-analyze and pre-cross-reference information it captured and collected.
Lanner’s LEC-2290, an IPC designed for intelligent edge computing, was eventually selected for this project. This box PC is powered by the 8th generation Intel® Core™ i7 CPU (codenamed Coffee Lake S) and features up to 32GB DDR4 memory, 4x PoE for camera connections and PCIe/Mini PCIe slots for GPU cards.
Live streams are captured using RTSP IP cameras connected to a 4G router and the LEC-2290 (where Gorilla IVAR™ is running) via PoE and/or RJ45 LAN ports, the captured streams, a.k.a. the inference data are then transmitted to the management site for data crunching. The LEC-2290 in this case leverages GPU-accelerated frameworks for speedy and latency-free data processing and video transcoding. Last but not least, the resulting information are pooled and archived for building a database on which further data mining can be carried out for even more accurate and precise analysis.
Edge AI-optimized traffic monitoring enables traffic flow analysis, vehicle counting, license plate recognition and driver/pedestrian behavior prediction, generating tremendous but useful volume of real-time data that can be relayed to the traffic control center for proactive responses with almost no latency, preventing infractions and potentially saving lives.
This project succeeded in illustrating not only the frequency by which traffic violation, such as illegal turns, occur, but also the capability Edge AI technology could deliver for aiding the law enforcement. Furthermore, this particular project elevated general public’s awareness for proper driving etiquette and pedestrian safety.
On average, a 50% decrease in violations were reported during the trial period; the data also demonstrates that during the trial period one of the most infraction-prone intersections recorded a monthly tally of 99 infractions, down from a monthly average of 307 infractions.
Such initial results not only look promising but also prove integrating tech solutions into urban planning and public safety is a worthwhile investment as all data and resulting analysis can be counted on for similar future projects in larger scopes and bigger cities.