Intelligent surveillance system

Abstract

A smart monitoring system mainly includes one or more front-end monitoring devices located in the front-end field and a back-end server system. The front-end monitoring devices run a front-end artificial intelligence model through a processing unit and have a key database. The front-end artificial intelligence model is used to identify key events in the monitoring data generated by the front-end monitoring devices. After identifying the key events, it transmits the monitoring data containing the key events to the back-end server system. The back-end server system then uses the back-end artificial intelligence model running in the back-end processing unit to identify the number of key objects in the monitoring data and confirms the key events a second time.

Claims

1. A smart monitoring system, comprising: One or more front-end monitoring devices, each front-end monitoring device comprising: a processing unit, wherein an object recognition module is implemented using artificial intelligence technology, the object recognition module being used to identify one or more objects in each frame of monitoring data and to determine a key object; and an event message module, the event message module generating and transmitting a message in response to a key event; an image capturing module, electrically connected to the processing unit, acquiring the monitoring data generated from a field captured by a camera lens; and a key database implemented using a storage device, electrically connected to the processing unit; and a back-end server system, connected to the one or more front-end monitoring devices via a network, comprising: A back-end processing unit includes an event receiving module implemented with artificial intelligence technology, used to receive the key event and corresponding monitoring data transmitted by one or more front-end monitoring devices after preliminary identification; a key object counting module used to calculate the number of objects in each frame of the monitoring data; an object identification module used with artificial intelligence technology to identify the one or more objects in the monitoring data, used to identify the behavior of each object and confirm the key event; and an event reporting module used to confirm the key event a second time, and generate an alarm message after confirming the key event a second time.

2. The intelligent monitoring system as described in claim 1, wherein the one or more front-end monitoring devices are one or more surveillance cameras located in a field, each being an edge computing device, to perform preliminary identification of the critical event.

3. The intelligent monitoring system as described in claim 2, wherein the monitoring data is real-time dynamic images captured by any of the one or more monitoring cameras, and a front-end artificial intelligence model running in the front-end monitoring device identifies one or more objects in the real-time dynamic images, and identifies the key event after comparing it with the key database.

4. The intelligent monitoring system as described in claim 3, wherein, In the front-end monitoring device, after one or more objects are identified by the front-end artificial intelligence model, an embedded vector of each object is calculated to compare with multiple key object vectors recorded in the key database, thereby identifying the key event that has one or more key objects.

5. The intelligent monitoring system as described in claim 3, wherein, The background server system obtains real-time dynamic images of the key event that have been initially identified by one or more surveillance cameras, runs a background artificial intelligence model to identify one or more objects in the real-time dynamic images, and confirms them as one or more key objects, thereby confirming the key event a second time.

6. The intelligent monitoring system as described in claim 1, wherein the key object counting module is a multimodal visual semantic model; when the background server system obtains the monitoring data containing the key event from the one or more front-end monitoring devices, the multimodal visual semantic model is used to describe the number of one or more objects in the monitoring data.

7. The intelligent monitoring system as described in claim 1, wherein a key event determination method is executed, the method comprising: The monitoring data is generated using one or more front-end monitoring devices in the intelligent monitoring system, wherein any one of the front-end monitoring devices runs a front-end artificial intelligence model. The front-end artificial intelligence model initially identifies the key event in the monitoring data; after identifying the key event, the monitoring data containing the key event is transmitted to the back-end server system; and a back-end artificial intelligence model identifies one or more key objects in the monitoring data to confirm the key event a second time.

8. The intelligent monitoring system as described in claim 7, wherein, In one of the front-end monitoring devices, a surveillance camera acquires real-time dynamic images of the same event from different perspectives, enabling the back-end server system to obtain the monitoring data from different locations via a network.

9. The intelligent monitoring system as described in claim 8, wherein the background artificial intelligence model running in the background server system is a multimodal visual semantic model. When the background server system obtains the monitoring data containing the key event from the one or more front-end monitoring devices, the multimodal visual semantic model is used to describe the quantity of the one or more objects in the monitoring data. When the quantity of the one or more objects is greater than 0, the system re-identifies whether the one or more objects are the one or more key objects.

10. The intelligent monitoring system as described in claim 9, which re-identifies the one or more key objects in the monitoring data, that is, confirms the key event a second time and issues a key event message.

Images