Enhanced Vehicle Detection
EVD, part of the Automated Order Taking (AOT) group at McDonald's, is an innovative independent project aimed at revolutionizing the drive-through experience. Our mission is to streamline and elevate the ordering process for every vehicle that enters our drive-through lanes.
Our Vision and Approach
At EVD, our key aim is to swiftly and effectively trigger a signal as soon as any vehicle, from cars to trucks to motorbikes, reaches the drive-through ordering system. This demands a mix of smart automation, state-of-the-art algorithms, and a synergy of IoT tools and advanced computer vision, all backed by robust engineering.
Deployment and Scope
The current iteration of EVD is operational in approximately 12 McDonald's restaurants across the United States, with a significant concentration in the vibrant city of Chicago. Our system's reach is set to expand, promising to impact more locations in the near future.
Object Detection Model using Computer Vision
Powered by YOLO3, our Object Detection Model identifies a range of critical components within a vehicle, including heads, arms, steering wheels, and driver-side mirrors. This accurate and efficient model serves as the foundation for the EVD system.
Components of Success: Project Solution
Our project solution comprises four key components, each contributing to the flawless functioning of EVD:
Random Forest Classifier
Building upon the output of our Object Detection Model, the Random Forest Classifier employs intricate algorithms to determine the readiness of a customer's vehicle. The results are categorized as either "True" or "False," indicating the preparedness of the vehicle for ordering.
Image Classifier using Convolutional Neural Networks
As a reliable backup, our Image Classifier uses advanced neural networks to assess vehicle readiness. This ensures smooth operations if the Object Detection Model faces issues. Outputs like "Ready," "Almost Ready," and "Not Ready" are interpreted by the EVD rules engine.
EVD Rules Engine
At the heart of our system lies the EVD Rules Engine, which intelligently processes signals from both the Object Detection Model and the Image Classifier. Through a series of logical rules, this engine makes decisive judgments about a vehicle's readiness. The resulting insights are then seamlessly conveyed to the API broker for further actions.
Precision in Hardware Systems
Camera - Evolution for Clarity
Presently, EVD operations rely on Vivotek Cameras within participating stores. However, we're committed to continuous enhancement. Our future plans involve transitioning to Axis Cameras, which offer superior performance, particularly in sun-glare situations. The enhanced image quality from Axis Cameras promises a sharper view, ensuring seamless performance throughout.
HME Loop Detector - Capturing Presence
Integral to our system is the HME Loop Detector, featuring an inductive coil beneath drive-thru lanes. This component generates signals upon the arrival or departure of a substantial object. This mechanism reinforces our commitment to accuracy and efficiency.