Harnessing constrained resources in service industry via video analytics

TL;DR

This paper demonstrates how integrating commercial video surveillance with deep learning enables real-time resource management and problem detection in service industries, exemplified by a trolley tracking system at Hong Kong International Airport.

Contribution

It introduces a practical system combining existing video technology with deep learning for resource optimization and customer service enhancement in the service industry.

Findings

High accuracy in detecting occlusions and resource issues

Significant improvement in daily operations

Potential for real-time customer assistance

Abstract

Service industries contribute significantly to many developed and developing - economies. As their business activities expand rapidly, many service companies struggle to maintain customer's satisfaction due to sluggish service response caused by resource shortages. Anticipating resource shortages and proffering solutions before they happen is an effective way of reducing the adverse effect on operations. However, this proactive approach is very expensive in terms of capacity and labor costs. Many companies fall into productivity conundrum as they fail to find sufficient strong arguments to justify the cost of a new technology yet cannot afford not to invest in new technologies to match up with competitors. The question is whether there is an innovative solution to maximally utilize available resources and drastically reduce the effect that the shortages of resources may cause yet achieving high level of service quality at a low cost. This work demonstrates with a practical analysis of a trolley tracking system we designed and deployed at Hong Kong International Airport (HKIA) on how video analytics helps achieve management's goal of satisfying customer's needs via real-time detection and prevention of problems they may encounter during the service consumption process using existing video technology rather than adopting new technologies. This paper presents the integration of commercial video surveillance system with deep learning algorithms for video analytics. We show that our system can provide accurate decision when faced with total or partial occlusion with high accuracy and it significantly improves daily operation. It is envisioned that this work will heighten the appreciation of integrative technologies for resource management within the service industries and as a measure for real-time customer assistance.