Tube Counter Accuracy on Two-Lane and Four-Lane Arterials

Objectives

1. Determine the accuracy of selected traffic counting devices in capturing actual vehicular flows.
2. Assess the accuracy of detection devices in classifying observed vehicular traffic.

The study compared three traffic data collection devices:

1. Jamar TRAX RD
2. MetroCount's MC5600
3. Traffic Tally 6 (TT-6)

Two test locations were used:

• A two-lane arterial (12th Avenue North, Fargo, ND)
• A four-lane arterial (45th Street, Fargo, ND)

Data was collected and compared against manual counts from video surveillance.

Key Findings
Two-Lane Arterial Results

Volume Results - Time Period 1
The surveillance video was reviewed from 11:00 a.m. to 1:00 p.m. to analyze vehicles traveling westbound on 12th Avenue. The results from the analysis are shown in Table 1

Both the Jamar and MetroCount devices produce very accurate results based on aggregate vehicle volume. The TRAX RD, MetroCount-1, and MetroCount-2 percentage differences were 1.5%, 0.0%, and 0.0%, respectively. The TT-6 device counted 31.6% more vehicles compared to the observed volume. The large difference in traffic volume may be attributed to heavy vehicles with several axles per vehicle. 
 

A comparison between the two MetroCount devices was performed to determine how multiple lanes affected the volume accuracy. Compared to the surveillance video, the MetroCount-1 (one lane) and MetroCount-2 (two lanes) counted 100% of the observed volume. Therefore, placing road tubes across both travel lanes on a two-lane road did not affect the volume accuracy.

The comparison between the Jamar TRAX RD and MetroCount-2 determined the volume accuracy of the devices when placing tubes across both travel lanes. Both devices provided accurate results, with differences of 1.5% for the Jamar device and 0.0% for the MetroCount device.

Classification Results - Time Period 1

Since the TT-6 device only consists of one tube, it cannot be used for the classification comparison.  Therefore, the comparison consisted of the Jamar device and the two MetroCount devices. Since classification devices have difficulty distinguishing between Vehicle Class 2 and 3, the net difference between the two classes was used for the comparisons. Table 2 illustrates the differences among the four categories of vehicles. Appendix C illustrates the differences for each of the 13-vehicle classes among the three devices.

Based on 326 vehicles, the Jamar, MetroCount-1, and MetroCount-2 devices incorrectly classified 12.6%, 8.0%, and 8.0% of the vehicles, respectively. Most of the classification errors were related to classifying single-unit trucks.
 

The MetroCount devices had slightly different classification inaccuracies. MetroCount-1 had 4 more errors related to passenger vehicles, while MetroCount-2 had 4 more errors related to single-unit trucks. 

Based on the inaccuracies of the single-unit trucks, both devices had problems related to Class 5 vehicles (two-axle, six-tire, single-unit trucks), as shown in Appendix C. 
When comparing the devices that spanned both travel lanes, the MetroCount-2 device was more accurate than the Jamar device. 

MetroCount-2 incorrectly classified 8.0% of the vehicles, while Jamar incorrectly classified 12.6% of the vehicles (note Appendix C). The Jamar device had more passenger vehicle errors (12 vehicles) but less error for trucks (6 vehicles). In addition, the Jamar unit counted 11 vehicles that were unable to be classified. When comparing the three truck categories, which had a total of 45 trucks, the Jamar and MetroCount-2 were incorrectly classified at 33.3% and 51.1%, respectively.
 

Volume Results - Time Period 2

After performing the first two-hour evaluation, a knot was tied into the Jamar road tubes and a new count was set up for the device. The knot allowed the device to collect data only for the westbound lane. The second time period gathered data from 4:00 p.m. to 6:00 p.m. Time period 2 was performed to compare the traffic data among the four devices (objective 1), the differences among the MetroCount devices (objective 3), and Jamar and MetroCount devices observing only one lane of traffic (objective 4).
 

Similar to time period 1, both the Jamar and MetroCount devices produced very accurate results. Based on 621 vehicles, the volume difference of the Jamar, MetroCount-1, MetroCount-2, were 0.0%, 0.0%, and 0.3%, respectively. The TT-6 device counted 14.5% more vehicles than the actual volume. This percentage difference is less than ½ of the recorded difference of time period 1, which was 31.6%. Although the volume of time period 2 (621) was approximately twice the volume of time period 1 (326), a higher percentage of trucks were observed in time period 1, resulting in more axle counts.
 

Similar to time period 1, the MetroCount-1 (one lane) and MetroCount-2 (two lane) devices were very accurate. MetroCount-1 had no difference (0.0%), while MetroCount-2 experienced a difference of 0.3%. Based on the one-lane evaluation, both the Jamar and MetroCount devices performed exceptionally well. Both the Jamar and MetroCount devices did not experience any difference (0.0%) compared to the actual volumes.

Classification Results - Time Period 2

This section compares the Jamar and MetroCount devices, which have tubes across only one of the two lanes on the bi-directional roadway. When comparing all of the vehicle classes, the Jamar device incorrectly classified 3.2%, while the MetroCount device incorrectly classified 1.9% of the vehicles (shown in Table 4). Therefore, this comparison shows that the devices are more accurate at classifying vehicles over one lane than two lanes on a bi-directional roadway.

Although the Jamar and MetroCount devices incorrectly classified four of the passenger vehicles, they displayed slight differences in the remaining vehicle classes (Appendix C). The Jamar device had the most difficulty classifying Vehicle Class 10 (six or more axle single-trailer trucks), with five errors.

 In addition, the device counted four vehicles that it could not classify. The vehicle class that provided the highest amount of error for MetroCount was Class 5 (two-axle, six-tire, single-unit trucks) with 3 vehicle count errors. Overall, the inaccuracies of the three truck categories for the Jamar and MetroCount devices were 26.6% and 17.8%. 

Four-Lane Arterial Results

Volume Results
All three detection devices counted less vehicles than the actual volume observed from the surveillance video (Table 5). Volume differences for the Jamar TRAX RD, MetroCount 5600, and the TT-6 were 9.5%, 3.0%, and 7.3%, respectively.

The lower traffic volumes can be attributed to vehicles simultaneously traveling over the road tubes in the two lanes, causing some vehicles to get skipped. Another reason for the discrepancies between the Jamar and MetroCount devices can be attributed to vehicle platoons arriving from the intersection of 12th Ave. North and 45th St., which is approximately one mile away from the test site. During numerous instances, vehicles would exhibit small gaps, which resulted in classifying two passenger vehicles as one heavy vehicle.

An additional factor that may have affected the performance of the Jamar TRAX RD relates to a “DBounce” feature of the detection device. The D-Bounce setting, which is the time the device will wait to recognize another vehicle, may have not been set low enough. The D-Bounce setting for the tests were 25 milliseconds (the default value) and could have been set between 1 and 100 milliseconds.

Classification Results

This section compares the vehicle classification results between the Jamar TRAX RD and MetroCount 5600 devices on a roadway with two lanes of travel per direction. When comparing all of the vehicle classes, the Jamar and MetroCount units incorrectly classified 35.0% and 6.3%, respectively (as shown in Table 6).

When comparing passenger vehicles, the Jamar device incorrectly classified 200 vehicles, while
MetroCount incorrectly classified 19 vehicles. The Jamar unit incorrectly classified 92 trucks (167.3%). Of the 92 trucks, Class 5 (two-axle, six-tire, single-unit trucks), Class 8 (four or fewer axle single-trailer trucks), and Class 1 (motorcycles) had vehicle differences of 53, 26, and 19 vehicles, respectively (as shown in Appendix C). 

The MetroCount device incorrectly classified 34 trucks (61.8%) and Class 5
provided almost half of the error with a difference of 16 vehicles.
As discussed in the previous section, potential discrepancies could be from a combination of vehicles simultaneously crossing over the road tubes and small gaps between vehicles. In addition, the classification analysis observed that both devices had difficulty distinguishing Class 5 (two-axle, six-tire single unit truck) vehicles. 

Conclusions

This comparative study of traffic data collection devices showcases the superior performance of the MetroCount tube counter across various testing scenarios. In both two-lane and four-lane arterial settings, the MetroCount device consistently demonstrated high accuracy in vehicle volume counting, with near-perfect results on the two-lane road (0.0% to 0.3% inaccuracy).

 Its classification accuracy also stood out, particularly in single-lane monitoring, where it misclassified only 1.9% of vehicles overall and a mere 0.7% of passenger vehicles. Even in more challenging multi-lane environments, the MC5600 maintained its edge, outperforming competitors in both volume and classification tasks. 

Notably, it showed significantly better truck classification accuracy than other devices, highlighting its versatility across vehicle types. While all devices faced increased challenges with multiple lanes, the MC5600 consistently proved to be the most reliable and accurate option, reinforcing its position as the leading choice for comprehensive traffic data collection needs.

For more detailed information on the study methodology and results, please refer to the full report by the Advanced Traffic Analysis Center at North Dakota State University.