Overview

更新时间:
复制 MD 格式

IoT Edge provides algorithm applications for configuring algorithms on video devices to process video data. It also provides algorithm services to manage algorithm tasks.

Background information

In the Internet of Things (IoT) industry, algorithm development often lacks unified definitions and standards. The high cost of in-house development is a major business pain point. To improve the efficiency of algorithm development and Operations and Maintenance (O&M), IoT Edge provides algorithm applications and services through V-series edge all-in-one devices. These devices support local analysis and inference for video and image data. Only necessary data is uploaded to the cloud, which reduces upstream bandwidth and traffic costs.

Glossary

Term

Description

Algorithm application

An application service provided by IoT Edge to configure algorithm parameters and view algorithm results.

Algorithm service

A service for retrieving, managing, and querying algorithm detection results. This service works with all algorithms.

Algorithm task

In an algorithm application, each enabled video device channel is considered one algorithm task.

Maximum number of tasks

The maximum number of algorithm tasks that can run concurrently on the current model of the edge all-in-one device.

Number of running tasks

The number of algorithm tasks that are currently running on the current model of the edge all-in-one device.

Algorithm applications

Four main types of algorithm applications are available. Each application includes one or more algorithm models to process or analyze data from video devices.

Note

Different V-series edge all-in-one devices support different types of algorithm applications. In the IoT Edge console, go to the Node Management > Algorithm Task Management page. Select your V-series edge all-in-one device to view the algorithm application types it supports.

Algorithm application

Included algorithm models

Scenarios

Facial recognition

Facial recognition

Suitable for facial recognition scenarios, such as at main entrances and exits.

People counting

People counting

Suitable for detecting the number of people passing through main entrances and channels. This is based on counting bodies or heads. It does not count the absolute number of unique individuals.

General object detection

  • Key area occupancy

  • Fire lane occupancy

  • Crowd gathering

  • Area intrusion

  • Trash detection

  • Hard hat detection

  • Absence from post detection

  • Flame detection

Suitable for scenarios that require setting detection targets and displaying the results.

Vehicle and license plate recognition

Medium-to-long-range license plate recognition

Suitable for scenarios that involve counting vehicles and recognizing license plates, such as at parking lot entrances and exits or for monitoring vehicle access in a campus.

Algorithm services

Algorithm services are basic services common to all V-series edge all-in-one devices. After you configure an algorithm application for a device, you can use algorithm services to manage the algorithm. Three algorithm services are available and are described in the following table.

Algorithm service

Description

Event Hub

Aggregates analysis event results from all algorithm tasks. It lets you query events from all algorithm tasks based on conditions such as time, video device, and algorithm application.

Retrieval service

Provides a retrieval service based on structured data, such as faces, human bodies, and vehicle license plates.

Base library service

Provides an image base library for the facial recognition algorithm. Use this service to upload, view, modify, and delete face images.

Important

The facial recognition algorithm application must use this service.

What to do next