Common methods of traffic counting
The intelligent transportation system market is seeing significant growth in the development of high-quality and real-time traffic information. One of the most widely used methods of gathering traffic data is through the use of road-based sensors, such as inductive loops. However, these traditional methods have drawbacks, including high costs for installation and maintenance, as well as limitations in detection capabilities. As a result, image processing-based solutions are becoming increasingly popular. In this article, we will introduce a new solution for traffic counting developed by Swarm Analytics and examine its effectiveness through an installation collaboration with SWARCO Schweiz AG in Switzerland. This investigation will focus on evaluating the solution’s accuracy for both counting and classification of traffic.
Swarm Perception Box installation
Swarm Analytics offers a convenient and easy-to-use solution for traffic data collection, which can be installed on existing infrastructure such as traffic light poles. This solution requires minimal setup and can be quickly installed on-site. In the case of the installation used in this article, SWARCO Schweiz AG in Switzerland simply ensured that power was available at the installation point and mounted the Swarm Outdoor Perception box, which includes a camera, on an existing traffic light pole. The camera was then calibrated using the Swarm Control Center, where various configuration options are available to be applied. This allows for a simple, efficient, and effective method of gathering traffic data without the need for extensive and costly installation and maintenance.
Data analysis of Swarm Analytics sensors
Our partner SWARCO Schweiz AG has implemented the Swarm solution in an installation that has been rigorously tested for data quality by comparing the results obtained from the Swarm Perception solution with those from a certified Swiss10 induction loop. This validation process involved manual counting of almost 2,000 vehicles and pedestrians.
As seen in the picture above, the installation includes a visual-based sensor represented by the Swarm Outdoor Perception box. This sensor captures real-time traffic counts and speed estimates, providing a comprehensive and accurate view of the traffic situation at the location at all times. The data is then visualized in the Data Analytics feature of the Swarm Perception Platform, which allows for easy monitoring and analysis of the traffic data.
Comparison of data
In this analysis, the data generated by the Swarm Analytics solution and the Swiss10 induction loop were compared to manual counts of ground truth across eight 15-minute time segments throughout the day, and under various traffic conditions. This comparison allowed an independent validation of the data. The analysis included a comparison of both traffic counts and vehicle classification (modal split) across the respective time periods. The manual ground truth data, which included a total of 1933 vehicles, ensured the validity and reliability of the analysis.
The results of this comparison indicate that both the Swarm Analytics solution and the Swiss10 induction loop are capable of providing accurate traffic counting data. However, it is worth noting that while the induction loop classifies vehicles according to the Swiss10 standard, the Swarm solution uses the standard classification from the BASt standard (8+1).
The counting accuracy at this location and time frame for the Swarm Analytics solution was found to be 97.0%, while the counting accuracy for the Swiss10 induction loop was 95.9%. Similarly, the average classification accuracy for the Swarm Analytics solution was 95.0% and for the Swiss10 induction loop was 93.9%.
Nevertheless, the Swarm Analytics solution exhibited a slightly higher level of stability in accuracy figures across different traffic situations, such as flowing traffic and slow-moving traffic. Additionally, the data revealed a clear correlation between the average vehicle speed and the accuracy of the inductive loop, with a decrease in accuracy at low average speeds, which typically occur during rush hours. This correlation was not observed in the Swarm Analytics solution.
Further information about Swarm solutions
Swarm Analytics provides a range of solutions beyond traffic counting. For gathering traffic insights in urban and inter-urban areas, Swarm Analytics offers real-time, 24/7 local traffic data generation with data analytics option. For more parking insights, the Swarm solution provides data on utilization, parking times for creating Parking Guidance Systems and analysis of parking spaces. Furthermore, there is an option to create even more advanced traffic insights across multiple installation locations, allowing the determination of travel times and an understanding of traffic flow in cities and municipalities.