Crowd Behaviour Analysis
Behaviour Analysis
Crowd
Crowd from a.i.tech is an application focusing on behavior analytics combined with a deep neural network capable of detecting people within the scene. Crowd functions include detecting overcrowding, social distancing, occupancy within a zone, dwell times, stationary people, people flow, loitering and more...
Functions
People Count
Count the number of people within one or more areas of interest and send a periodic notification containing such information. .
Occupancy of Zone
Assess the percentage of occupancy within one or more areas of interest, providing a periodic notification containing this information or an instant notification if a user-configured threshold is exceeded.
Loitering
Verify the presence of people loitering in a region of interest, sending an alarm if the person wanders too much time in the area. The maximum loitering time is configurable by the user.
Where did they go?
Send the trajectory of every person when leaving the scene. .
Overcrowding
Check for overcrowding situations by sending a notification whenever the number of people in one or more areas of interest exceeds a user-configurable threshold. .
Dwell Time
Compute the dwell time of a person in a region of interest, by sending a notification containing this information for each person detected in the area. .
People Flow
Count the number of people crossing one or more virtual lines, sending a notification for each crossing detected. .
Camera requirements
All the functionalities are available with frontal or plan view. We recommend for social distance verification, that a frontal view is used. .
Social Distancing
Verify social distance compliance by sending an alert if the people in a group are deemed too close. Both the minimum number of people in the group and the minimum distance between two people are user-configurable.
Stationary People
Verify the presence of people steady in a region of interest, sending an alarm if the person remains stationary for too much time. Both the maximum latency time and the radius within which the person is considered steady are configurable by the user.
Origin Matrix
Count the number of people crossing one area and arriving in another, building the origin- destination matrix. The number of areas, the type (origin and/or destination), the position within the scene and their shape can be configured by the user.
People counting gate
In the case where the goal is to count people passing through a gate entrance we would recommend using the application "People" The application requires a plan view and is optimized for this scenario.
Integrations
The events generated by the video analysis applications can be notified by e-mail, with the possibility of attaching a sequence of images with the metadata related to the event of interest.
FTP
Notifications can be sent to FTP servers (e.g. FileZilla, IIS), storing for each one a CSV file and a sequence of images with the metadata relating to the event of interest.
HTTP
A.I. Tech offers maximum integration with standard HTTP and HTTPS services. Notifications can be received by customizing requests to REST or CGI services to control remote devices (e.g. for PTZ handover mechanisms) or in JSON format, attaching one or more images to the event of interest on any third-party server.
UDP /TCP
The metadata associated with the events generated by the A.I. Tech can be received via text strings over configurable UDP / TCP protocols, such as those sent by POS systems.
MODBUS
A.I. Tech plugins can forward alarm notifications using the Modbus TCP protocol, according to the specifications outlined by the IEC 61158 Standard.
MQTT
A.I. Tech plugins can forward notifications using the publish-subscribe MQTT protocol, according to the specifications outlined by the ISO/IEC PRF 20922 standard.
ONBOARD
A.I. Tech also offers mechanisms to notify events of interest directly on board the device that processes the images. It is in fact possible to activate, through applications, the digital outputs on the AI-Appliance, the digital output of the Hanwha cameras and the notification mechanisms of the Axis cameras.
System Architecture
Eco System
Products from A.I. Tech are technologically compatible with the following products
Common Functions
of A.I. Tech Apps
Tampering
If your camera does not have this functionality, don't worry. All apps can generate when tampering events occur.
PTZ Handover
Do you need to control a PTZ camera when an event occurs? With the ability to trigger a sequence of HTTP calls, you can bring the camera to the desired pre-set, start the PTZ tracking (if available on the camera) and bring the PTZ camera back to the home position after a configurable amount of time.
Activation on a time basis
Apps can be activated on a time basis, only in specific time intervals, e.g. from 9 p.m to
8 a.m. In addition, for server or embedded apps, it is also possible to specify the day (e.g. Monday from 10am to 12pm., Tuesday from 6pm to 10pm.).
Masking
In all apps masking of moving objects can be activated, in order to be compliant with all kinds of privacy requirements.
AI-Dash-Embedded
This is the app's built-in dashboard, which allows you to have the data for the specific camera at your fingertips at all times ( even on camera, if you have an SD card).
​
​
​
Dynamic activation via software
In addition to time-scheduling, you can activate your apps dynamically, through software calls, via HTTP GET requests that allow you to operate in two modes: impulsive and level-based. In the first case you have a single START request and the app will remain active for a certain predefined amount of time. In the second case, you have one START request and one STOP request to control the app.
Use Cases
MUSEUMS
Occupancy of zones, dwell times
PEDESTRIAN ZONES
Identifying overcrowding and social distancing
GYMS
Occupany on machines, overcrowding
TRANSPORT
Loitering on platforms, People flow
Ready to get in touch ?
Contact Us
