Opportunity
Circle K required a way to inform customers about vehicle queue lengths without retaining or publicly streaming the underlying camera images. They approached BMK (Atea Group), an IT solutions provider, for an automated, privacy-focused alternative.
Challenge
- Data Minimisation: Avoiding the storage and public streaming of identifiable camera images.
- Variable Conditions: Diverse setups across multiple facilities, including:
- Different camera models and varying image quality.
- Varied camera placements and angles.
- Challenging weather conditions (sunlight, rain, snow) affecting accuracy.
Solution
BMK collaborated with 1D.works to deploy a sophisticated, privacy-focused computer vision system:
- AI-Powered Vehicle Counting: Used deep-learning algorithms to estimate the number of cars without exposing the source images through the customer-facing service.
- Adaptive System Training: Provided configurable parameters for facility-specific conditions and a path for evaluation with new data.
- Real-Time Data Integration: Transmitted queue-length information through an API integrated into Circle K’s customer-facing website.
Delivered Capabilities
- Privacy-Focused Processing: Designed the customer-facing workflow around queue counts rather than stored or streamed imagery.
- Queue Data API: Supplied current queue estimates to the existing website through an API.
- Site Configuration: Allowed parameters to be adjusted for different camera placements and environmental conditions.
- Evaluation Path: Made it possible to test and refine model performance as representative data became available.
The architecture reduced the amount of identifiable imagery exposed by the earlier workflow. It should not be read as a legal determination of GDPR compliance or as evidence of customer-satisfaction, usage, or cost outcomes.
Explore our AI Workflow Integration service. Read: Private and Secure AI.