Comparing video-based traffic counting to pneumatic tubes

 

To compare the performance of pneumatic road tubes with video-based traffic counting, Traffic Databank conducted a 72-hour study spanning seven intersections in New York City. This head-to-head comparison conclusively demonstrated that video-based traffic counting is superior in many areas, including:

  • Technician safety: Installing and retrieving road tubes required technicians to step into harm’s way for extended periods of time, often during especially hazardous conditions; in contrast, installing and retrieving each Miovision Scout (video collection unit) took only a few minutes and did not require technicians to enter the roadway.
  • Data Accuracy: Road tubes suffered from chronic overcounting (by 14%) and acute undercounting (by 35%), severely harming the integrity of the traffic study data; Miovision Scout suffered no accuracy problems and delivered accurate real-world data.
  • Auditability: Ordinarily, there is no way to audit road tube data — suspected erroneous data must be discarded, and unnoticed errors corrupt the dataset; Miovision Scout provides a completely auditable video record (and was used in this case to audit the road tube data and diagnose the reasons for inaccurate counts).
  • Cost-effectiveness: Installing and removing the road tubes was six times more expensive — and considerably more dangerous — than performing the same work with Miovision Scout.

Setting up an objective head-to-head comparison

To create an objective comparison, Traffic Databank installed Miovision Scout units and JAMAR ATDR pneumatic tubes to conduct a 72-hour traffic study at seven busy intersections in New York City (four in Manhattan and three in Brooklyn).

Since the Scout collects video-based data, Traffic Databank was able to use the recorded footage as a source of ground truth for all counting comparisons.

Keeping technicians out of harm’s way

Despite safety strategies and guidelines that nearly all US states have implemented, according to the Federal Highway Administration (FHWA) fatalities from work zone crashes are all too common. Any time a technician physically enters the roadway to install or retrieve traffic count equipment, that individual is at serious risk.

In this study, due to extremely high vehicle congestion throughout the day, technicians were forced to install and retrieve the pneumatic tube equipment between the hours of 10 p.m. and 2 a.m. During these low-visibility hours, the technicians were exposed to 650 vehicles during the average 30-minute installation period.

In contrast, installing the Miovision Scout units took an average of 10 minutes; most importantly, installing, removing, and servicing Scout units is performed without the technician entering the roadway. With tools like the Scout, that provide reliable video traffic counts, it is an unnecessary — and unacceptable — risk to expose technicians to the dangers of pneumatic tube installation and retrieval.

Providing a reliable foundation for traffic engineering

Road tubes record when the passage of vehicle axles generates an air impulse. These impulses are then interpreted to count traffic movements. Tube installations can be configured to provide different outputs (e.g., classification, speed) by using multiple tubes spaced apart.

Video classification, like that provided by Miovision, uses video of the study location as the foundation for traffic analysis. Today’s systems use artificial intelligence to examine the video, classifying vehicles and generating turning counts with great accuracy — and with complete auditability.

Even though road tubes have been a hallmark of the industry for decades, Traffic Databank’s study demonstrated beyond a doubt that video provides vastly superior accuracy. Auditing the data revealed many reasons for this performance difference.

In many common circumstances, tubes overcount traffic. Comparing the tube counts to the validated Miovision counts showed that tubes overcounted vehicles by 13% during the study period. Auditing the video revealed the main causes to be high congestion, high volume, low speeds, or the presence of freight vehicles (i.e., with more than two axles). Each of these factors drastically impacted the interpretation of pneumatic air impulses.

ny-customer-story-graph2

As shown in the graph above, pneumatic tubes (blue) tended to overcount vehicle traffic compared to the ground truth (red).

Equipment failure and other factors impair count reliability

Equipment failures (e.g., breakages), communication problems, and the inability to secure road tubes properly throughout the duration of counts are all factors that lower the quality of data collected.

The inability to remotely communicate with the tube equipment means that when the collector becomes damaged or disconnected, this impairment doesn’t become known until the units are retrieved or a technician physically visits the collection location.

Even within this relatively short study, the road tube installations suffered from impairments that caused severe undercounting of traffic:

  • A street maintenance vehicle passed through the location and physically disconnected the pneumatic tube
  • On several occasions, vehicles parked on the tubes, rendering them useless—sometimes for hours at a time
    In contrast, the Miovision Scout units suffered no such difficulties. If they had, then technicians could investigate remotely; if a physical site visit was needed, it still wouldn’t require technicians to step into the roadway.

ny-customer-story-graph1

As shown in the line graph above, pneumatic tubes (blue) suffer from frequent failure, dedicated by the sudden and large dips in vehicle counts.

Tubes degrade rapidly and are sensitive to weather

Both weather and degradation have the potential to impair traffic counts, especially in longer study periods. Although the 72-hour study period benefited from favorable weather conditions, it is well-known that road tube performance suffers in inclement weather. While this study was too short to experience problems due to general wear-and-tear (notwithstanding the maintenance vehicle that ripped the tube out of its counter), road tubes tend to degrade rapidly under heavy traffic flow.

Auditing traffic data to ensure accurate counts

To assess the accuracy, completeness, and validity of traffic data from pneumatic tube equipment, algorithms search for anomalies and flag potentially erroneous data for manual review. Unfortunately, since there is no source of ground truth to verify the data, there is not much recourse except to discard the data in question (whether erroneous or not). Slow-and-steady errors, like the overcounting noted above, will not be flagged.

In stark contrast, video-based traffic studies generate an auditable record (i.e., the video itself). In fact, it is only through the Miovision Scout video that the performance of the road tubes in this study could be objectively assessed and the reasons for inaccurate counts could be diagnosed.

ny-customer-story-graph-screenshot

With Miovision, video can be audited to ensure accuracy—whether to examine unusual findings or to perform random “spot-checks”.

Understanding road tube misconception

One of the most common arguments in favor of road tubes is their perceived cost-effectiveness. But is that really true? While pneumatic tubes can be used to provide a quick method of performing spot studies or rough traffic counts at a small number of locations (subject to the inaccuracies noted previously), larger studies along corridors with multiple busy intersections prove especially challenging.

Due to the invasive nature of installing the pneumatic tube equipment directly in the roadway, efficient installation and retrieval of the pneumatic tube equipment is next to impossible for larger studies. Plus, installation and retrieval during heavy periods takes longer, which is why road tube work is often performed in the overnight hours.

Although the primary objective of this study was to compare the performance of road tubes versus video-based traffic counts, the project nevertheless produced compelling evidence that undermines the perception that road tubes are cost-effective. Operationally, installing and removing the road tubes was six times more expensive:

  • Miovision Scout units were installed in an average of less than 10 minutes and can be installed by a single technician; removal took even less time
  • Installing road tubes took two technicians an average of 30 minutes

“Think about completing a 50-intersection study in this city. At two technicians per intersection, I’d need nearly 100 people to complete the installation and retrieval of the count equipment.”

— Oumar Barro, Traffic Databank