The IndiGO solution from Bluecity features lidar sensors and artificial intelligence to capture realtime
multimodal traffic data. Recently, the City of Edmonton collaborated with Bluecity to evaluate the
potential of using its lidar andAI technology for future safety planning initiatives.
IndiGO differentiates itself by using lidar sensors combined with an AI perception software that
provides real-time multimodal traffic data in any weather or lighting condition, all while preserving the
privacy of road users.
As intersection collisions cause nearly 70% of all traffic injuries, the City of Edmonton had made the
installation of automated traffic-safety technology at intersections a priority. Safety Devices (ISDs)
monitor intersections and photograph vehicles that speed or run through red lights. ISDs have been
reducing red-light running and speeding, thereby lowering the number and severity of crashes.
ISDs have been helping Edmonton move towards its goal of zero fatalities and major injuries. With
lidar and AI technology now available for traffic data collection, the city planned to evaluate the
accuracy of the traffic count and vehicle speed of the lidar data
Bluecity’s IndiGO, a lidar and AI solution, was
installed at the southwest corner of 104 Avenue
and 109 Street intersection. While it was
recommended that to obtain the best results,
two lidar sensors should be installed in the
intersection, the city opted for one sensor,
recognizing that this would be pushing the limits
of its capabilities.
The collected lidar data was processed and
transmitted to Bluecity’s online dashboard
(IndiGO iQ). IndiGO iQ provides metrics
including traffic counts, average vehicle speed
and a conflict map, as shown in Figure 1. The
daily traffic counts include pedestrian, bicyclist,
bus, passenger vehicle and truck, by
intersection turning movements. The speed
chart shows average speed by intersection
approach. The conflict map allows users to
specify the turning movement, road user type
and conflict threshold, and download lidar video
footage for the conflicts.
The study compared the accuracy of the traffic
count and vehicle speed of the lidar data with
manually collected data. The data was collected
on June 17, 2021, from 3 pm to 5 pm. Bluecity
would manually collect the data.
The lidar data would also be compared to the
ISD data. To validate the similarity between the
manual and ISD traffic count data, ISD data
was compared to the manual data collected by
Bluecity. A difference of only -1.8% was
determined, so it was concluded that manual
count and ISD data were very close, and their
counts validated each other.
Results of manual vs lidar traffic count comparison
During the two-hour period, the lidar counted 7,583 vehicles compared to 7,601 vehicles that were counted manually, representing an overall difference of only -0.2%. In terms of absolute percentage errors, the north and west approaches had 1%, while the south and east approaches were slightly higher (1.6% and 2.1%).
When counting turning movements, the counts for right-turning vehicles at the south and east approaches had higher absolute error percentages (24% and 27%). This could result from the size of the intersection and not having enough lidar resolution at the far corner. This could be easily remedied by adding a second lidar sensor.
Results of speed data comparison
The City of Edmonton was also interested in the accuracy of detecting the speed of vehicles. For the speed comparison, manual speed data was collected using the distance measured from the Google Satellite Distance Measurement feature, and timestamps when vehicles entered and left the intersection area that was bordered by the stopping lines (see Figure 2). Only the through vehicles were tracked. In total, 393 vehicles were manually tracked.
Overall, the lidar data had an error percentage of only 1% and an absolute error percentage of 3.1%. The north approach had the highest error percentage of 3.3%. In terms of the actual values, it ranges from -0.2 to 1.5 km/h for the speed difference and 1.1 to 1.8 km/h for the absolute speed difference.
Results of Manual vs Lidar traffic count comparison
With these impressive results, the city wanted to see how the lidar performed when compared to intersections already equipped with other traffic enforcement technology. During the period from May 15 to June 13, 2021, the daily traffic counts using the lidar technology were compared to the ISD data.
The average daily difference is 1.1% for the north-south and north-west directions. In the case of the east-west traffic, the average daily difference rose to -8.5%. This increase could be due to the distance from the lidar sensor as the results align with the comparison between lidar sensors and manual counting where the study showed -5% for east-west traffic and only .5% for north-south and north-west traffic.
In addition, the study showed that the differences are larger during weekdays when traffic volume is higher, than during weekends when traffic volume is lower. This was attributed to the use of only one lidar sensor which caused occlusion
Results of Lidar vs ISD speed data comparison
Data was collected during the same time period, from May 15 to June 13, 2021, however, some ISD data was not validated and some lidar data was missing.
Both of these were not included in the evaluation but were not expected to impact the results.The study showed the average difference to be 1.3 km/h (3.6%) for the east-west traffic and -0.2 km/h (-0.4%) for the north-south and north-west traffic.
Since the ISD measures vehicle speed using loops that are located before the stopping line, while the lidar speed data were measured within the intersection, the ISD data include vehicles decelerating and accelerating due to the traffic signal change.
A further comparison was conducted using the same date period but that only looked at vehicles moving 50 km/h and faster. The average difference is 0.4 km/h (0.7%) for the east-west traffic and 0.7 km/h (1.3%) for the north-south and north-west traffic.
As was done with the traffic count data, the 50 km/h and faster vehicle speeds were further broken down into hours. The vehicle speeds are similar between lidar and ISD data across the hours with the average hourly difference only 0.3 km/h (0.5%) for the east-west traffic and 0.7 km/h (1.4%) for the north-south and north-west traffic.
The City of Edmonton was pleased that the l idar data provided excellent accuracy in terms of traffic count and vehicle speed data even when only one sensor is used when compared with 2 hours of manual data, the traffic count difference is only -1.4% to 0.7% at approach level and the average speed difference is only -0.4% to 3.3%.
When compared with 30 days of ISD data, the average daily traffic count difference is -8.5% to 1.1% at turning movement level and the average daily speed difference is -0.4% to 3.6% or 0.7% to 1.3% for vehicle speeds at 50 km/h or higher.
The study showed that the traffic count is more accurate when the approach is closer to the lidar equipment, which could be due to occlusion. The traffic count accuracy for right turning is lower than that of left turning and through traffic but this could be easily addressed by installing additional Lidar sensors. In addition, although the average speed is very close between lidar and ISD data, the maximum speed measurement for lidar data is lower than that of ISD data.
These results also align with previous studies performed by Bluecity. The City of Edmonton is pleased to be able to continue its safety planning for the upcoming years knowing that Bluecity IndiGO solution can provide accurate, real-time, multi-modal data and many advanced safety metrics.
The Bluecity team is also committed to providing additional features for traffic safety, such as a variety of Automated Traffic SignalPerformance Measures (ATSPM), pedestrian and driver behaviour analysis, safety metrics summaries and more.
Bluecity’s goal is to make streets safer for pedestrians, cyclists and motorists. We envision a world where traffic accidents are rare, driving to work is a pleasure and carbon emissions are reduced. Make sure to read our follow-up articles and see how IndiGO’s accuracy continues to improve – coming soon.