top of page

SHELL PENNSYLVANIA CHEMICALS PLANT

We supported Shell Polymers’ construction megaproject in Pennsylvania with a tailored solution that automatically detected inaccuracies in the as-built documentation by comparing it against the geo-spatial data developed from drone imagery.
In this project we verified the accuracy of the as-built underground (UG) layouts by comparing the drawings against TBs of drone imagery taken over the duration of the works to date.
​
The site was 390 acres, and data acquisition was with high resolution drone imagery, captured over 2.5 years divided into 120 giant (tiled) images.
​
large_images.png

ACQUIRE IMAGERY

Existing and new data, imagery and construction drawings are captured in real-time for rapid analysis and comparison.

IMAGE ANALYSIS

Unsupervised Machine Learning automatically generates thousands of labels for faster and improved analysis.

UNDERSTANDING THE IMAGE

Unsupervised Machine Learning automatically generates thousands of labels for faster and improved analysis.

Picture4_edited.jpg
The underground firewater pipes and manholes are detected in the drone images using segmentation algorithms. 

The most recent reliable geo-spatial time-stamp is selected to conduct a comparison with the CAD documentation and detect the deviation of the installed component from the initial designed drawing.
 
Once the CAD data is pre-processed in the right format and the desired layers are isolated, the system is straightforward and runs quickly.

The system is able to autonomously detect areas of issue and to highlight them in a nice and efficient way for manual review by the end user.

The system can be scaled to an area of any size, as well as being able to scaled to evaluate the positions of other underground components available in CAD. 
bottom of page