Multi-Camera Groundplot Tracking for Cashier-Less Retail Checkout
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cashier-less checkout systems in retail stores rely on depth cameras or LIDAR for customer tracking, which require excessive GPU resources and can lose tracking IDs if a camera malfunctions, leading to incorrect billing and customer dissatisfaction.
Innovation Solution
A system using media acquisition devices like cameras positioned to cover multiple points in a physical space, projecting input data points onto a groundplot, and assigning tracking IDs through clustering, reducing resource consumption and ensuring accurate tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth cameras or LIDAR systems are used for groundplot tracking to detect and track customers, then object detection capability is improved, but GPU resource consumption increases excessively
Solution Approach 1:
The system segments the detection task by using multiple media acquisition devices positioned at different locations, where each device covers a specific zone. This distributes the computational load across multiple devices rather than requiring one high-performance system to process entire space data, reducing overall GPU resource consumption while maintaining detection precision.
Solution Approach 2:
The system transitions from traditional 3D depth-based tracking to 2D groundplot projection tracking. By projecting 3D spatial data onto a 2D groundplane and tracking objects based on their projected positions, the system reduces computational complexity and GPU resource requirements while maintaining accurate object detection and tracking capabilities.
2Device complexity
If a single depth camera or LIDAR system is used for tracking, then device complexity is reduced, but tracking reliability decreases when the device malfunctions
Solution Approach 1:
Each media acquisition device is positioned to cover a specific local zone with optimized detection parameters for that area. This local specialization allows each device to operate independently and reliably within its coverage area, and the system maintains overall tracking reliability by aggregating data from multiple local devices, so if one fails, others continue to track their respective zones.
Solution Approach 2:
The system implements redundancy by deploying multiple media acquisition devices throughout the space. This creates a backup capacity where if one device malfunctions or stops operating, other devices continue to provide detection and tracking coverage, preventing loss of tracking ID and maintaining system reliability without requiring complex failover mechanisms.
3Reliability
If multiple media acquisition devices are deployed to cover the entire space, then tracking reliability is improved, but device complexity increases
Solution Approach 1:
Multiple media acquisition devices perform the same universal function of detecting and tracking objects within their respective coverage zones. Each device uses identical detection algorithms and tracking methods, allowing the system to scale by simply adding more devices without increasing operational complexity. The groundplot projection approach provides a universal coordinate system that simplifies integrating data from multiple devices.
Solution Approach 2:
The system uses identical media acquisition devices positioned at different locations, each performing the same detection and tracking functions. This copying approach allows reliable coverage of the entire space through replication of proven working units, reducing the complexity of designing and managing heterogeneous systems while maintaining high tracking reliability through redundant coverage.
Data Source
AI summary
The improved systems and methods utilize a detection and tracking device in communication with a plurality of media acquisition devices strategically placed within a defined space, which is divided in a grid fashion such that each media acquisition device senses at least three points within the space, with adjacent media acquisition devices overlap, such that the plurality covers the entire space. Objects in the space are sensed by respective media acquisition devices, annotated with a tracking identifier, and tracked while moving therein by either the media acquisition devices or the detection and tracking device. The detection and tracking device derives input data points from each media acquisition device for projection onto a groundplot of the space, clusters the input data points from corresponding media acquisition devices into a single cluster, assigns the tracking identifier to a centroid of the cluster, and tracks the object as it moves.