Image source «Photobank Lori»
We strive for excellence - that is why we pay special attention to collection of relevant input data.Today there are more and more opportunities to automate traffic monitoring. The means to determine the main parameter – traffic volume - have long been the subject of development in Russia and abroad.
This year we tested a range of devices available on the market in an effort to improve the technology we use. There are two main solutions: a radar method and machine vision.
The first method is generally used at multi-lane highways with a radar placed at 90 degrees to the roadway. Measurement data are transferred online to the analytics system and are immediately ready for use. The second one uses video cameras above or on the side of the roadway and involves data postprocessing that takes time comparable to the duration of the shooting.
To assess the capabilities and efficiency of the various systems, we tested the accuracy of vehicle type recognition based on dimension estimation, as well as detection of speed and number of vehicles over a certain period.
Some of the comparative tests were carried out during diagnostics of the federal highway on the St. Petersburg Flood Protection Barrier.
We installed radio frequency counters to record traffic volume and composition at the same point. Each type of counters had similar settings and direction angle.
In addition, we installed a video camera to use machine vision and verify measurement data by manual checking of the videos.
Comparison and analysis of the data obtained confirmed the accuracy declared by the manufacturers of different counters (90-95%). In general, the systems showed almost identical results. The differences were in the methods and ease of installation and configuration.
It would not be right to call these results definitive, since we cannot claim that we have covered all options, and the technologies are constantly improving. However, we can name Wavetronix SmartSensor systems (USA) and Houston Radar SpeedLane Pro (USA) as the most acceptable types of radio frequency radars.
As for machine vision systems, we can note two software vendors – Czech DataFromSky and Russian TrafficData. Each has its own advantages. In addition to vehicle classification these systems detect the speed and recognize number plates (the latter allows for vehicle trajectory modelling); counting of pedestrians is also available. Shooting from drones can be an advantage if the problem of a short battery life is solved. This was discussed in detail by Igor Shamirov at the Transport Planning and Modelling conference.
Conclusion: both methods reduce labor input and human error, but require significant investments in equipment, software, and staff training.
Having mastered these technologies, we are happy to apply them even to small projects.
The first method is generally used at multi-lane highways with a radar placed at 90 degrees to the roadway. Measurement data are transferred online to the analytics system and are immediately ready for use. The second one uses video cameras above or on the side of the roadway and involves data postprocessing that takes time comparable to the duration of the shooting.
To assess the capabilities and efficiency of the various systems, we tested the accuracy of vehicle type recognition based on dimension estimation, as well as detection of speed and number of vehicles over a certain period.
Some of the comparative tests were carried out during diagnostics of the federal highway on the St. Petersburg Flood Protection Barrier.
We installed radio frequency counters to record traffic volume and composition at the same point. Each type of counters had similar settings and direction angle.
In addition, we installed a video camera to use machine vision and verify measurement data by manual checking of the videos.
Comparison and analysis of the data obtained confirmed the accuracy declared by the manufacturers of different counters (90-95%). In general, the systems showed almost identical results. The differences were in the methods and ease of installation and configuration.
It would not be right to call these results definitive, since we cannot claim that we have covered all options, and the technologies are constantly improving. However, we can name Wavetronix SmartSensor systems (USA) and Houston Radar SpeedLane Pro (USA) as the most acceptable types of radio frequency radars.
As for machine vision systems, we can note two software vendors – Czech DataFromSky and Russian TrafficData. Each has its own advantages. In addition to vehicle classification these systems detect the speed and recognize number plates (the latter allows for vehicle trajectory modelling); counting of pedestrians is also available. Shooting from drones can be an advantage if the problem of a short battery life is solved. This was discussed in detail by Igor Shamirov at the Transport Planning and Modelling conference.
Conclusion: both methods reduce labor input and human error, but require significant investments in equipment, software, and staff training.
Having mastered these technologies, we are happy to apply them even to small projects.