The AI middle platform is to modularize artificial intelligence technologies such as deep learning, computer vision, knowledge graphs, and natural language understanding, and to integrate artificial intelligence capabilities such as hardware computing capabilities, algorithm training capabilities, model deployment capabilities, and basic business display capabilities. , Combined with the data resources of the middle platform, it is packaged into an overall middle platform system.
In the EasyCVR video sharing fusion cloud platform, the AI middle platform is a safety production early warning platform that provides artificial intelligence video analysis services. It consists of an artificial intelligence reasoning center, an algorithm market, an artificial intelligence computing service center, and an early warning center, forming a structured management model of "1 market and 3 centers". By providing flexible and convenient artificial intelligence analysis services, the AI middle platform can accelerate the intelligent production, management and operation of various industries, support industry management decision-making and business development, and further improve the efficiency of intelligence.
The structured management model of the AI middle station includes the following parts:
1. Artificial Intelligence Computing Service Center: Responsible for unified scheduling and management of hardware resources and algorithm resources, making full use of the computing power of computer CPUs, customized data streams, and reasoning chips, and realizing the configuration of multiple algorithms and multiple cameras, so as to achieve efficient AI computing operations .
2. Artificial intelligence reasoning center: responsible for algorithm call, analysis deployment, open API and scene recognition data output. It adopts a mid-platform design to manage and call algorithm analysis services, algorithm SDKs and AI reasoning chips downwards, and provide open APIs, application plug-ins and data output capabilities upwards to achieve fast and standardized AI video analysis and platform integration.
3. Early warning center: Supports customization of early warning conditions according to different application scenarios, automatically saves image data when the early warning conditions are met, and pushes relevant platform systems to quickly report alarm information. At the same time, fine-grained data control and management are realized through the permission assignment function.
4. AI algorithm market: manage AI algorithm models for all scenarios, including but not limited to algorithm model version management, model file storage management, algorithm model upgrade, algorithm model dynamic expansion, algorithm model call, model addition and deletion management, scene model authority management wait.
In the EasyCVR platform, the AI platform has scene recognition capabilities by default, including cluttered material detection, area intrusion detection, tripwire intrusion detection, smoking detection, crowd density detection, smoke and flame detection, personnel departure detection, helmet wearing detection, Wearing work clothes detection, vacant sleeping post recognition, fire exit occupancy detection and other functions.
AI middle platform is widely used in actual scenarios. The following are some application cases:
1. Smart transportation: By analyzing traffic video data, timely discover abnormal events such as congestion, accidents, road damage, construction, natural disasters, and illegal events on urban and intercity traffic roads, improve traffic coordination and management, and improve traffic safety and fluency.
2. Passenger flow management: Accurately count the number and density of passenger flow in scenic spots, shopping malls, stations, airports and other areas, and identify abnormal events such as early warning fights, crowd gatherings, etc., to meet large-scale, high-density passenger flow management and safety assurance needs.
3. Urban environmental governance: Automatically identify urban management issues such as road occupation, exposed garbage, motor vehicle occupation of sidewalks, forest fire prevention, etc., capture relevant issues, and realize case reporting.
4. Production/service management: For work scenes with special requirements such as service windows, factories, construction sites, etc., identify abnormal behaviors such as personnel wearing clothes, leaving and sleeping on duty, automatically obtain evidence and trigger early warnings, and realize automation and intelligence of job management.
EasyCVR platform has the following technical advantages:
The EasyCVR platform supports the access of multiple protocols and device types, as long as the device supports standard protocols (such as national standard GB/T28181, RTMP, RTSP/Onvif protocol, etc.) or manufacturer's private protocol (such as Hikvision/Dahua SDK, Hikvision Ehome protocol etc.), can be quickly and easily accessed to the platform. It is not limited to device types, brands and models, and can cover almost all video source devices on the market, such as IPC, NVR, encoders, video servers, drones, mobile soldiers, emergency control balls, patrol terminals, etc. . The platform can quickly connect to front-end equipment in various regions, gather massive video resources to the platform for intelligent identification and analysis, surveillance video viewing, capture, recording and storage, video stream distribution, alarm message push, video data sharing and other services, providing flat The video aggregation and fusion management capabilities are personalized, visualized, and comprehensive.
Based on advanced technologies such as video surveillance technology, network transmission technology, GIS geographic information technology and AI image intelligent recognition technology, the EasyCVR platform aggregates, manages, intelligently analyzes, processes and distributes the transmission data of all front-end sensing devices, and pushes early warnings, enabling fast, Accurately discover potential safety hazards and give early warning. It realizes all-round and three-dimensional monitoring of video, and provides support for video monitoring related units and departments to establish a scientific, sound and effective intelligent safety production and life monitoring and early warning system.
