The water system of commercial/park/residential water system park projects is widely distributed, and manpower inspection cannot achieve full-time + full-range coverage, especially in the nighttime green area, which has great hidden dangers in management, and the existing monitoring system lacks effective reminder/alarm means. The automatic inspection platform with the function of "full-time + full-range + accurate identification + timely reminder/alarm" jointly developed by Hckdigi Technology and Shanghai Deep Silicon realizes the effective management of the project's water system and park.

1. The demand for intelligent management of water system parks

Compared with conventional parks, the management of existing water system parks has the following pain points:

• Manpower Patrol: Manual patrol with long intervals

• Centralized monitoring: monitoring large screen manpower tracking and monitoring

• Night Patrols: There are patrol intervals and time blind spots

• Video storage: Flipping through and looking for it is time-consuming and labor-intensive

• Artificial slackness: inability to concentrate all the time, negligence and omission

• Architectural style: bridge and tunnel/building/greening and other monitoring blind spots

• River system: There is no parameter setting for the water system monitoring camera

• Real-time alarm: timely reminder to upgrade the sound and light situation

• Abnormal events: identify/monitor/full-time overlooks

• Evidence solidification: Video/image evidence of abnormal events is preserved

The intelligent management solution for water system parks is oriented to various scenarios in the park, providing functions including safety management and intelligent supervision. In the safety management scenario, through video supervision, event warning, system docking and other empowerment, the construction site safety management and protection from passive response and manual disposal to intelligent processing and real-time alarm have improved management efficiency.

2. Smart video management platform

The intelligent management platform of water system parks is based on the main equipment access capabilities, AI capabilities and related management capabilities. All kinds of modularization are divided to realize free and flexible functions at the application level of actual scenarios, and ensure centralized and unified management at the top level at the management level of the overall system. Whether it is a single-platform system or a multi-platform cascading system, it can meet the needs of system management and platform capability use.

The platform features are as follows

· Basic platform client: web management interface, mobile APP applet

· Home dashboard: provides a summary page of various indicator information, and you can browse a brief summary of the items of interest in real time

· Real-time monitoring: Provide video streams gathered by front-end monitoring terminal devices, which is convenient for real-time monitoring and meets the needs of large-screen point display

· AI alarm: Aggregate alarm information related to AI algorithm hardware, alarm screenshots, alarm clip videos, and manage by permission, organizational structure, and alarm type. After the algorithm is triggered, the illegal processing needs to be delivered to the task system (personnel management, regional management, equipment management in the area under its jurisdiction, alarm push, push processing, and processing list management), and at the same time it is delivered to the IP network (docking with the IP network system), which can be controlled in conjunction with other management subsystems and devices

· Data report statistics: provide equipment and online information, AI alarm statistics, and big data analysis. Through the background management software and early warning information database, the classification and summary of early warning events are realized, and the early warning events of the same kind and the same region are classified. Generate daily, weekly, monthly, quarterly, and annual reports to provide decision-making analysis

· System settings: Provide basic information configuration and user-related permission management such as company/team management, device management, rule configuration, and recognition mode management

· Device management: manage camera access, binding, and unbinding

· Algorithm allocation: The specific scene algorithm required by the corresponding algorithm can be assigned to the connected camera separately

· Open API: provides basic information configuration and API three-party push interface system

· Call and mobile communication management: centralized management of mobile phones, smart phones, smart servers, and inspection recorders

· Algorithm library: face recognition, trajectory tracking, motor vehicle/non-motor vehicle/pedestrian/site monitoring, major disasters, object recognition, voice conversion, text recognition and other types of algorithms to meet the needs of different scenarios

 · The platform solution can be used as the central node for the access, management, distribution, analysis and processing of surveillance video data, and the stream-based architecture can effectively meet the needs of various scenarios and video AI algorithm analysis. Among them, the signal source includes video signals, such as ONVIF protocol mainstream brand cameras, GB/T28181 protocol, EHOME protocol cameras, as well as the connection of drones, law enforcement recorders, IP sound columns and other equipment

AI intelligent computing gateway

Hckdigi Technology's Silicon Gravel series AI intelligent computing gateway is based on Nvidia Jetson Nano/NX and Orin Nano/NX Series/AGX Orin core modules, pre-installed Ubuntu operating system, providing 20/40/70/100/200/275Tops high computing power. The device supports USB, CAN bus, GbE, MIPI camera interface FPC connector, M.2 Key M interface expansion SSD storage, large-capacity SATA hard disk storage, and supports 4/5G and WIFI wireless communication. The equipment adopts a strong and lightweight aluminum alloy material design and a high-efficiency passive heat dissipation design, with high performance, low power consumption, strong environmental adaptability, and can operate stably in harsh industrial environments. Through the combination of diversified deep learning algorithms, video structuring, face recognition, behavior analysis, status monitoring and other applications can be realized, and the equipment can be used in the Internet of Things, intelligent manufacturing, autonomous driving, smart home, smart security, smart medical and other fields.

3. AI intelligent management platform algorithm

4.The advantages of AI intelligent management platform

1) Compatible with existing camer

In view of the situation that conventional monitoring equipment and systems have been deployed, it supports different brands such as Hikvision, Dahua, and Uniview, and supports 2K, 4K, and 8K high-definition video, data collection, transmission, and interaction, and through the integration of public and private protocols of different brands of cameras, the video is re-encoded and decoded through the front-end hardware, and then transcoded into a private protocol on the platform, and can be packaged into any video stream format required by customers, such as m3U8, FLV, RTSP, RTMP, H5 live broadcast link for algorithm analysis platform and third-party supervision platform to call.

It supports cross-network segment, multi-brand, non-fixed IP camera access, high-concurrency live streaming platform for audio and video streaming media, opens up the protocol layer of video sources, and is compatible with multi-brand IPC cameras, NVR, DVR video streams, GB protocol access, unified video aggregation and distribution, forwarding, and cascade of higher-level supervision platforms.

2) Algorithm customization and tuning of the self-developed platform

The entire platform architecture and software are independently developed by the company, and the underlying algorithm framework is deeply integrated and developed based on mainstream deep learning frameworks such as Pytorch, which supports a variety of different algorithm models such as face recognition, trajectory tracking, motor vehicle/non-motor vehicle/pedestrian/site monitoring, major disasters, object recognition, voice conversion, text recognition, etc., and can be integrated into a unified framework to meet the needs of full-scene video AI analysis, and provide customers with algorithm customization and tuning services for multiple special scenarios.

3) Web terminal + mobile App + SMS alarm push

In order to improve the response ability of the system, the alarm information and screen situation of the monitoring point are notified to the management and disposal personnel through the web terminal.

The platform can also provide API interfaces to connect with other information application platforms, push alarm information through mobile phone applets or SMS, and grasp the real-time situation of monitoring points through mobile apps anytime and anywhere in the process of disposal, so as to improve emergency response capabilities and effectively shorten emergency response time.

5. Shanghai Panlong Tiandi AI intelligent early warning system