CorVS: Person Identification via Video Trajectory-Sensor Correspondence in a Real-World Warehouse

2025-10-30

AI Summary

In real-world warehouse environments, robust person identification solely from visual trajectories remains challenging due to occlusions, appearance variations, and sensor noise. To address this, we propose CorVS—a novel method that jointly models the correspondence probability and reliability between video-based trajectories and wearable-sensor time-series data, incorporating temporal consistency constraints for high-accuracy identity matching. CorVS employs a deep learning model to predict association probabilities between trajectory and sensor segments, and integrates these predictions into a multi-object temporal data association framework for identity inference. Evaluated on a large-scale real warehouse dataset, CorVS achieves significant improvements in both accuracy and stability over state-of-the-art methods, demonstrating superior robustness against realistic disturbances and enhanced practical applicability for industrial deployment.

Technology Category

• Intelligent Robots: State Estimation
• Computer Vision: Motion & Tracking
• Planning, Routing, and Scheduling: Activity and Plan Recognition

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Worker location data is key to higher productivity in industrial sites. Cameras are a promising tool for localization in logistics warehouses since they also offer valuable environmental contexts such as package status. However, identifying individuals with only visual data is often impractical. Accordingly, several prior studies identified people in videos by comparing their trajectories and wearable sensor measurements. While this approach has advantages such as independence from appearance, the existing methods may break down under real-world conditions. To overcome this challenge, we propose CorVS, a novel data-driven person identification method based on correspondence between visual tracking trajectories and sensor measurements. Firstly, our deep learning model predicts correspondence probabilities and reliabilities for every pair of a trajectory and sensor measurements. Secondly, our algorithm matches the trajectories and sensor measurements over time using the predicted probabilities and reliabilities. We developed a dataset with actual warehouse operations and demonstrated the method's effectiveness for real-world applications.

Problem

• Identifying individuals using video trajectories and sensor data correspondence
• Overcoming limitations of visual-only person identification in warehouses
• Matching worker trajectories with sensor measurements for reliable tracking

Innovation

• Deep learning predicts trajectory-sensor correspondence probabilities
• Algorithm matches trajectories with sensor measurements over time
• Method tested with actual warehouse operations dataset