Lightweight Intelligent Monitoring Platform Architecture and Component Overview

Modern intelligent monitoring platforms are often feature‑rich but become bulky, hindering private deployment; a lightweight solution is therefore needed.

01. What is WVP? Web Video Platform (WVP) implements the GB28181‑2016 standard, providing core signaling, device management, NAT traversal, and supports IPC/NVR from brands such as Hikvision, Dahua, and Uniview, as well as standard‑compliant cascading.

02. Intelligent monitoring products based on WVP By rebuilding on the open‑source WVP framework, powerful business functions are combined with its robust video foundation to create a new, lighter, more comprehensive, and faster‑to‑deploy monitoring platform.

03. Web server technical architecture

Component responsibilities and functions:

Camera: provides video streams (GB‑standard cameras must register to SIP server).

SIP Server: handles authentication, registration, heartbeat, stream control for GB cameras.

ZLMediaKit: C++11 high‑performance media server offering various streaming protocols.

KubeEdge: open‑source edge‑computing platform that pulls streams from MediaKit, extracts frames, invokes AI, and pushes events to Kafka.

AI Service: provides face recognition, intrusion detection, smoking detection, etc.

OSS: stores AI event images/videos uploaded by KubeEdge.

Kafka: message queue between KubeEdge and the web backend.

Web Server: offers device, skill, event management, statistics and analytics.

MongoDB: stores cloud and AI events.

MySQL: stores web‑server data.

Milvus: stores facial feature vectors and supports similarity search.

04. Camera skill issuance sequence diagram

The diagram shows how AI skills are assigned to cameras.

Frontend sends camera‑skill configuration as a JSON string to the web server.

Web server stores the info in the database and forwards it to the edge side, which consumes Kafka AI events.

KubeEdge pulls streams from MediaKit, extracts frames, calls AI interfaces, and uploads detected events to S3 and Kafka.

S3 stores images and short videos and provides download URLs.

Kafka delivers events to the web server for consumption.

05. GB‑standard camera SIP registration flowchart

Pre‑assign camera username and password.

Camera backend fills SIP server IP, credentials, GB ID, channel ID, etc.

Camera sends registration request to the SIP address.

SIP server replies 401 with authentication challenge.

Camera resends registration with credentials and receives 200 OK.

This uses Digest authentication, common for GB‑standard cameras.

06. Real‑time streaming architecture for GB cameras

Sequence of SIP INVITE/200 OK/ACK messages establishing media sessions among the media stream sender, SIP server, and media server, followed by B2BUA proxying and media flow.

07. Feature showcase

Overview page:

Core functions include multi‑protocol support (Hikvision, RTSP, GB), skill configuration and issuance, alarm and analytics visualization, device management, map positioning, cloud replay, live playback, license‑plate database, and face‑library management.

08. Technical innovations

Replacing MySQL with MongoDB for event storage reduces write latency from ~90 ms to ~10 ms, handling up to ~2 million daily events.

Using KubeEdge for flexible edge pod scheduling eliminates manual camera‑box binding and improves load handling.

Separating image and short‑video uploads accelerates alert timeliness.