Deepening Edge Computing Applications to Support ICBC’s Digital Transformation

ICBC actively tracks the development dynamics of edge computing technology in the industry, exploring the architecture, key technologies, and capability construction of edge-cloud systems for the Internet of Things (IoT). Through continuous validation and practice, ICBC has taken the lead in the industry to establish an enterprise-level edge-cloud collaborative service system, forming a core capability for edge-cloud collaboration that covers multiple dimensions including resources, data, and intelligence, providing users with offline, low-latency, and highly flexible edge computing services. Currently, the edge-cloud collaborative service system has managed over a hundred edge computing nodes, accumulating nearly 20 edge intelligent atomic services such as customer operation detection and speech quality inspection, providing technical support for the rapid implementation of various edge computing scenarios in the financial sector.

1. Establish an Integrated Resource Collaboration and Scheduling Capability. Following the model of “edge-end resources, cloud management,” based on a cloud-native edge computing framework, ICBC has built an integrated edge-cloud resource collaboration scheduling and management capability. The edge computing framework provides a containerized deployment environment for edge applications, shielding different hardware differences of devices. Edge application development only needs to focus on business functionality implementation without worrying about the underlying hardware differences of edge nodes, simplifying the development difficulty of edge applications. Meanwhile, the cloud IoT platform achieves unified management and scheduling of computing, storage, and network resources of edge nodes through the management of edge gateways and edge integrated machines, providing on-demand allocation, elastic expansion, and dynamic scaling resource scheduling services for edge applications, supporting rapid iteration and launch of innovative business scenarios.

2. Build Data Analysis and Processing Capabilities for Edge-Cloud Collaboration. Following the principle of “near storage, full excavation,” ICBC has constructed data analysis and processing capabilities for edge-cloud collaboration. For low-value density source data, it is stored nearby on edge computing devices, while data with potential value is uploaded to the cloud for unified storage and analysis in a data lake, fully excavating the enormous value contained in the bank’s unstructured data. Edge nodes are mainly responsible for the collection of on-site/terminal data, performing preliminary processing and analysis based on rules or data models, and uploading high-business-value relevant data and processing results to the cloud. The cloud is responsible for aggregating and integrating massive, multi-source data for analysis, excavating the enormous value contained in the data to support innovative business applications.

3. Build Support Capabilities for Edge Intelligent Model Development. Following the principle of “cloud training, edge-end inference,” ICBC has built an edge intelligent model development production line based on the capabilities of in-house IoT, artificial intelligence, big data, and other cloud platforms, lowering the technical threshold for edge intelligent application development and meeting the rapid launch and promotion needs of edge intelligent application scenarios in the bank. Edge/terminal generates training data uploaded to the cloud. The cloud quickly builds specialized domain models based on data annotation, model construction, model training, and evaluation tools. According to the hardware conditions and business scenario requirements of edge nodes, the model conversion and compression tools are used to convert the models built in the cloud into edge intelligent models, which are then pushed to the edge for deployment, achieving a closed-loop process of data collection, model training, and model deployment.

Edge computing provides sufficient computing power nearby, integrating artificial intelligence algorithms to offer edge intelligent services at the network edge, meeting the critical needs of the financial industry’s digital transformation in real-time response, reducing transmission costs, and excavating unstructured data, providing better support for financial business scenarios that require high real-time performance and bandwidth, and enhancing the quality and efficiency of financial services.

1. Build a New Model for Intelligent Risk Control to Enhance Risk Prevention Capabilities. Utilizing edge computing and computer vision technology, a new model for risk intelligent management, “AI-assisted review + human re-examination,” is constructed to achieve real-time, comprehensive monitoring of branch business transactions and key locations, providing timely alerts and intervention for potential risks, ensuring the safety and compliance of financial services throughout the process.

Taking dual recording quality inspection in the wealth management room as an example, video segmentation services and intelligent dual recording quality inspection models are deployed on branch edge computing devices. The recordings are segmented into several video clips based on scenario types such as “product explanation, customer response, action operation, information verification,” and processed for speech recognition, semantic understanding, document detection, and action recognition for each video segment. This effectively solves the pain points of complex dual recording compliance inspection processes, time-consuming manual quality inspection, and inconsistent standards, transforming post-inspection into real-time inspection, enhancing the timeliness of dual recording quality inspection, and protecting customers’ legitimate rights and interests.

For example, in the agile security warning of branches, based on edge computing, suspicious item detection, and abnormal event detection technologies, real-time analysis of security video 24/7 is achieved. It identifies abnormal behaviors such as falls, fainting, sleeping, gathering, group events, and charging of electric vehicles in self-service banks and branches, generating alert information to remind monitoring center personnel to pay attention and handle it properly, effectively preventing the escalation of incidents, transforming from “post-event evidence collection” to “real-time alerts,” enhancing proactive prevention capabilities and case handling efficiency.

2. Innovate Online and Offline Collaborative Service Models to Improve Customer Service Experience. By integrating audio and video applications and intelligent video analysis technologies, ICBC transforms branch edge computing into an online and offline interactive hub, strengthening online and offline connection services, breaking down barriers between online and offline channels, and building collaborative service capabilities that link online scheduling with offline services, meeting real-time, localized business service needs, and innovating customer service models at branches.

For example, in the accompanying service at branches, video analysis services such as ReID and target tracking are deployed on edge computing devices to monitor each customer’s waiting time in the lobby. When a customer’s waiting time exceeds a threshold, a notification is immediately sent to the lobby customer service manager. When the lobby customer service manager is busy, they can issue a command to the smart robot using a Pad to provide care services such as tea and conversation to the customer, forming a group collaborative service model centered around the lobby manager and smart robot, helping to create a financial service experience that feels more technological, futuristic, and warm.

For example, in the holographic video call scenario at branches, based on edge computing and other technologies, an intelligent engine for image quality is created to process audio and video data for high-fidelity compression and restoration, achieving functions such as super-resolution enhancement of video quality. By establishing a holographic video call channel between branches and remote agents, it breaks the limitations of traditional video calls in two dimensions, achieving real-time two-way interaction in naked-eye 3D images, providing customers with an immersive “face-to-face” communication experience. When customers engage in holographic video calls with remote business experts, the holographic pod application supported by edge computing makes it feel as if the remote business expert is “sitting” in front of the customer, providing professional investment and financial consultation, product recommendations, and business handling guidance, thus enhancing marketing and service at branches.

3. Build a New Model for Branch Operation Management to Enhance Intelligent Operation Levels. Leveraging edge computing and computer vision technology, ICBC creates an intelligent operation system for business halls by deploying cameras and edge computing power at branches, achieving real-time intelligent analysis of video monitoring data locally, intuitively reflecting the operation status of business halls, effectively enhancing the refinement level of branch management, promoting rational scheduling of various resources at branches, and releasing human resources to improve marketing conversion rates.

For example, in the branch portrait screen, based on the computing power infrastructure of edge devices at branches, technologies such as target detection, attribute recognition, and trajectory tracking are comprehensively applied to conduct real-time analysis and processing of video streams at branches. By recognizing attributes such as gender and age of customers arriving at the branch and promptly pushing this information to customer service managers, it provides auxiliary means for precise marketing. At the same time, it tracks the trajectory of customers arriving at the branch, achieving functions such as customer flow statistics, analysis of customer stay duration, real-time queue numbers at service windows, and area density heat analysis. This data is presented in real-time to branch management personnel, helping them gain insights into customer flow, area personnel entry and exit, and density congestion, assisting in optimizing business processes and rationally adjusting business area distribution, self-service equipment placement, and staffing.

For example, in the non-site service supervision of branches, technologies such as cross-camera tracking and human posture estimation are applied to conduct localized analysis and processing of video streams at branches, achieving intelligent supervision checks of branch employee service behaviors, replacing manual inspections of the environment and service processes. It actively identifies non-compliant behaviors such as unauthorized assistance and operating without customers present, promptly reminding branches to optimize and improve, enhancing the service level of branch employees, thus improving customer experience and satisfaction.

In the future, ICBC will strengthen in-depth exploration of industry standards, scenario applications, and financial ecosystems related to edge computing, continuously promoting the upgrade of financial services and the transformation of business models, and comprehensively advancing the construction of a digital ICBC and the development of the edge computing industry.

First, promote the construction of edge computing application standards. Based on ICBC’s successful practical experience in planning, route selection, custom development, and application adaptation of edge computing systems, it will collaborate with various parties in the edge computing industry to deeply excavate and refine standardized proposals, solidly advancing the formulation of edge computing application standards in the financial industry, laying a foundation for the healthy development of the edge computing industry. Second, expand edge computing applications. Based on the current situation and future development needs of financial business, using the edge-cloud collaborative service system as a foundation, integrating technologies such as artificial intelligence, IoT, audio and video, and robotics, continuously deepening the “edge computing +” integrated technology innovation applications, further strengthening the depth and breadth of edge computing applications in the financial industry, and promoting the quality and efficiency of financial services. Third, strengthen financial ecosystem construction. Around the problems faced by corporate clients in their digital transformation process, provide corporate clients with a comprehensive service solution of “finance + technology,” and promote high-quality development of enterprises by formulating differentiated service strategies to provide more precise digital services for corporate clients, further expanding the financial ecosystem.

(This article was published in the “Financial Electronics” August 2024 issue.)