By Wayne Lee and Luke Gervais, BCIT Geomatics Supported by McElhanney
Camera-based photogrammetric mobile mapping platforms are increasingly being considered for infrastructure asset mapping, particularly where field teams need to capture larger areas more efficiently while maintaining reliable positional accuracy. For survey and geomatics professionals, the key question is not whether these systems can collect data quickly. The more important question is whether they can produce results that meet professional expectations when compared with established survey methods.
A field performance evaluation conducted at the British Columbia Institute of Technology Burnaby campus assessed Looq AI’s Platform, a handheld multi-camera photogrammetric mapping platform that combines imagery, GNSS, and inertial sensing to produce georeferenced 3D data, for infrastructure asset management. The study compared Looq Platform results against Total Station and GNSS RTK survey methods, with particular attention to feature extraction accuracy, wet surface performance, and the role of external ground control in high-accuracy workflows.
The study focused on three principal questions.
First, whether the Looq Platform could detect and position manhole and utility cap features with accuracy comparable to Total Station and GNSS RTK survey methods.
Second, whether surface moisture would affect the geometry or detection accuracy of the photogrammetric model.
Third, how external ground control supports professional-grade positioning when using AI-enabled reality capture for infrastructure mapping.
The findings indicate that the Looq Platform can support precision infrastructure asset mapping when used with rigorous survey control protocols. The results also reinforce a familiar survey principle: geodetic-grade applications benefit from a strong external control framework.
Figure 1. A field operator uses Looq AI’s qCam to capture the BCIT study area during the infrastructure asset-mapping evaluation.
Study Area and Reference Framework
The evaluation was conducted within a 150-meter by 50-meter outdoor plaza on the BCIT Burnaby campus. The site included hardscaping, mature trees, overhead canopy structures, and utility features. A pre-established Total Station control network tied to the campus geodetic monument network provided the absolute reference framework for the evaluation.
Figure 2. Aerial view of the BCIT Burnaby campus study area showing surveyed utility features, control locations, and reference targets used in the field evaluation.
The study area contained approximately 45 to 47 targeted manhole covers and utility features. Spatial data was collected through four measurement approaches.
A Leica TS13 Robotic Total Station was used to establish the reference coordinates for the study, while a Leica GS06 GNSS RTK receiver provided a comparison.
The Looq Platform was evaluated using a sunny, dry-condition capture processed with localized survey control and a post-rain, wet-condition capture. The rainy control and no-control datasets were generated from the same raw capture session; the processed results were downloaded once before and once after applying the platform’s “align to controls” function. Because the raw capture data was identical, this comparison isolates the effect of control alignment on absolute positioning without introducing additional field-capture variability.
All coordinate data was processed and compared in UTM Zone 10N. A nearest-neighbor spatial matching process was used to pair detected features with the Total Station reference points. Residuals were computed in Easting and Northing, and statistical significance testing was conducted at α = 0.05.
Because the acquisition sessions were not conducted simultaneously, the feature sets did not overlap perfectly across every dataset. The study notes that differences in matched point counts reflect session-to-session coverage variation rather than systematic bias.
Total Station and GNSS RTK
The Total Station survey served as the absolute ground truth for all comparative analyses. A closed traverse using the Leica TS13 was tied into the existing BCIT geodetic monument network through three localized control points. A Leica round prism was used to establish the control points and traverse stations, while a Leica mini prism was used specifically to survey the manhole and utility features. The resulting observations were adjusted using MicroSurvey STAR*NET least squares software.
The adjustment processed 114 observations over seven stations and passed the Chi-square statistical validity test, confirming internal consistency of the control network.
The GNSS RTK dataset was collected using a Leica GS06 receiver through the Leica SmartNet CORS network. Control points were occupied for a minimum epoch of 30 seconds, while discrete infrastructure features were occupied for three to five seconds.
The site presented common GNSS challenges. Dense canopy and nearby building facades intermittently affected satellite geometry. Some features located directly beneath tree canopy could not be occupied due to insufficient satellite availability and were omitted from the GNSS dataset.
After cleaning the GNSS RTK dataset by removing five outlier points, the GNSS RTK benchmark achieved 25.7 mm 2D RMSE against the Total Station reference.
Looq Platform Capture Workflow
Each capture session began by establishing a network connection to the Leica SmartNet CORS service. This allowed the system to acquire an initial GNSS fix for trajectory seeding.
Each capture session was conducted as a 15-to-16-minute continuous pedestrian trajectory through the study area. Raw multi-camera image and IMU sensor data was uploaded to the Looq processing environment, with a typical processing turnaround of 12 to 24 hours.
For the control-based workflows, trajectory optimization and absolute georeferencing were achieved by incorporating Total Station-derived control point coordinates into the processing pipeline. The resulting point cloud was exported as a high-resolution orthorectified GeoTIFF.
Manhole cover features were then identified in Autodesk Civil 3D. Feature centroids and cover diameters were extracted to CSV using the Civil 3D DATAEXTRACTION command. Statistical analysis was performed in Python 3 using SciPy.
Figure 3. Point-cloud output generated from the Looq capture of the BCIT campus study area, showing buildings, vegetation, walkways, and site features within the test area.
Positional Accuracy with Control
When constrained by ground control, the Looq Platform achieved strong positional results against the Total Station benchmark.
The Looq sunny-condition control dataset achieved 13.0 mm 2D RMSE. The Looq rainy-condition control dataset achieved 14.8 mm 2D RMSE. By comparison, the cleaned GNSS RTK dataset achieved 25.7 mm 2D RMSE.
The control-based Looq datasets detected 39 to 42 of the 45 to 47 surveyed targets, corresponding to 87 to 89 percent recall. Approximately five additional detections appeared in the Looq data but were absent from the Total Station survey. Because both the sunny-condition and rainy-condition datasets agreed on these additional detections, the study interprets them as likely real infrastructure features not collected by the Total Station survey rather than false positives.
The results support the conclusion that, when used with ground control, the Looq Platform can produce positional accuracy suitable for precision infrastructure asset mapping within the conditions tested.
Wet Surface Performance
The study also evaluated whether wet pavement affected Looq Platform performance. To test this, the sunny dry-surface and rainy wet-surface control-adjusted datasets were compared using 37 common matched points.
The results showed no statistically significant difference in positional accuracy between the sunny and rainy condition datasets. The mean positional difference between conditions was approximately 1.1 mm and was not significant.
Diameter accuracy also showed no statistically significant difference between conditions. The mean diameter difference was approximately 2.0 mm and was within measurement noise.
All statistical tests for both position and diameter comparisons produced p-values greater than 0.33. Within the level of precipitation represented in the study, wet surface conditions did not meaningfully degrade positional or diameter accuracy.
Field observations added useful context. Manhole covers with shallow standing water showed reduced surface texture in the point cloud because the specular water layer affected the visibility of the cover pattern. Even so, the circular geometry remained visible in the observed cases. The report notes that deeper standing water could potentially obscure the visible boundary of a cover, which presents an opportunity for additional evaluation under more extreme conditions.
Figure 4. A wet manhole cover photographed after rainfall. Standing water reduced visible surface texture, while the circular feature boundary remained identifiable in the tested conditions.
Ground Control and Professional Survey Workflow
One of the most valuable operational findings was the role of ground control in achieving survey-grade accuracy.
The Looq rainy-condition control dataset achieved 14.8 mm 2D RMSE. In the no-control test conducted for this study, the dataset showed 85.8 mm 2D RMSE against the localized Total Station reference. These values came from two processed outputs of the same rainy-condition raw capture: one downloaded before and one after applying “align to controls.” This difference highlights the value of localized control when the objective is geodetic-grade accuracy. The same-capture comparison shows that local feature geometry remained consistent while localized control improved agreement with the site reference framework.
Importantly, the no-control dataset maintained local feature consistency. The primary difference was a coordinate-frame offset rather than a loss of relative geometry or feature detection performance. In practical terms, the Looq Platform preserved the internal structure of the mapped features, while external control provided the survey-grade anchor needed for absolute positioning.
This distinction is valuable for infrastructure teams because it reinforces a familiar survey principle: high-accuracy mapping depends not only on the capture technology, but also on the control framework used to georeference the data. When paired with localized ground control, the Looq Platform demonstrated strong accuracy for precision infrastructure asset mapping.
The report notes that both the control-adjusted and no-control datasets were processed using the same UTM Zone 10N coordinate reference system and the same Looq processing pipeline. The no-control dataset showed a consistent northward displacement of approximately 84 mm, while maintaining relative geometry. The study did not determine the cause of this offset. A processing or workflow error remains one possible explanation, but the source was not isolated through additional testing. Additional evaluation could help further isolate the source of the offset, including the potential influence of the initial GNSS seed or processing workflow.
Diameter Measurement Considerations
Diameter extraction was also evaluated as part of the study. The Looq Platform measurements showed a modest tendency to report manhole cover diameters below the field reference values, with observed differences influenced by both field measurement and extraction methods.
The report notes that the reference measurements were collected manually by two operators using a tape measure and recorded to the nearest 5 mm. Looq-derived diameter values were also manually extracted in Autodesk Civil 3D by fitting circles to features from the orthorectified GeoTIFF. Because both methods involved operator judgment, the observed differences likely reflect a combination of measurement, interpretation, and workflow variability.
The study observed an average tendency for the extracted diameter values to be approximately 8 to 13 mm smaller than the manual field measurements. Because both the field measurements and the Civil 3D extraction involved operator judgment, further testing would be required before applying a standardized correction. For similar workflows, teams may wish to compare a sample of extracted dimensions against field measurements as part of their quality-control process.
Operational Implications
The evaluation offers several practical takeaways for infrastructure asset mapping.
With ground control, the Looq Platform achieved 13.0 to 14.8 mm 2D RMSE against Total Station ground truth. This was stronger than the cleaned GNSS RTK benchmark of 25.7 mm within the study area.
Wet surface conditions did not produce a statistically significant reduction in positional or diameter accuracy under the conditions tested.
The system matched 87 to 89 percent of Total Station-surveyed targets, and the additional unmatched Looq detections were likely real features absent from the Total Station dataset.
GNSS RTK produced five outliers, representing 16 percent of the points collected. Several were attributed to field data management factors. A comparable data quality issue was not observed in the Looq datasets.
In the no-control test conducted for this study, the Looq dataset exceeded the 50 mm tolerance referenced in the report for GIS asset management. The result demonstrates that localized ground control substantially improved absolute agreement with the site reference framework in this test, rather than establishing expected performance for every no-control workflow.
Study Scope and Future Evaluation
As with any field evaluation, the findings should be interpreted within the scope of the conditions tested.
The weather comparison was based on one sunny-condition capture session and one rainy-condition capture session, with 37 paired points. Within those conditions, the results showed no statistically significant difference between wet and dry surfaces. Additional evaluations across a broader range of weather events would further strengthen the weather-resilience conclusion.
Diameter measurements also included expected workflow variability. Both the field reference measurements and the Civil 3D extraction process involved manual judgment, which likely contributed to some of the observed differences.
The no-control dataset showed a consistent georeferencing offset while maintaining local feature consistency. Because the control-adjusted and no-control results were derived from the same rainy-condition raw capture and the cause of the offset was not isolated, the result should be interpreted as specific to the test configuration evaluated rather than as a universal characteristic of no-control Looq processing. Further evaluation could help isolate the source of the offset and determine how the workflow performs across different control, datum, epoch, and GNSS initialization conditions.
Overall, these considerations do not diminish the study’s primary finding. When used with rigorous survey control protocols, the Looq Platform demonstrated strong potential for precision infrastructure asset mapping in real field conditions.
Conclusion
This field evaluation shows that camera-based photogrammetric mobile mapping can be a viable tool for precision infrastructure asset mapping when used with rigorous survey protocols.
When constrained by external ground control, the Looq Platform achieved 13.0 to 14.8 mm 2D RMSE against Total Station ground truth and outperformed the cleaned GNSS RTK benchmark in the study area. The platform also demonstrated resilience to wet surface conditions, with no statistically significant effect on positional or diameter accuracy under the precipitation conditions tested.
At the same time, the study reinforces the importance of sound survey practice for geodetic-grade infrastructure mapping. In the no-control test conducted for this study, the dataset maintained local feature consistency but showed an absolute georeferencing offset. Because the cause of the offset was not determined, the finding should be understood as evidence that localized ground control improved absolute positioning in this test, not as a definitive measure of all no-control performance.
For geomatics professionals, the central lesson is that camera-based photogrammetric mobile mapping can expand field mapping workflows while complementing established survey methods. When paired with localized ground control, careful processing, and appropriate quality checks, the Looq Platform demonstrated strong potential for infrastructure asset mapping in real field conditions.
