Alarm condition personnel dynamic association system and method based on multi-mode identification

Abstract

The invention relates to an alarm personnel dynamic association system and method based on multi-modal recognition, and the method comprises the steps: collecting personnel face information through a face base library module, inputting the information into a face base library, configuring an intelligent analysis task, adding a camera and a channel in an equipment management module, and adding the intelligent analysis task to select a corresponding to-be-analyzed channel, and the intelligent analysis server pulls the video stream, synchronously performs face trajectory tracking and behavior analysis, matches a face trajectory ID in an alarm condition coordinate range according to a detected behavior alarm condition, and if the associated trajectory ID in an alarm area does not identify a specific person when the alarm condition occurs, the intelligent analysis server performs face tracking and behavior analysis. The dynamic updating module marks a unique key and pushes an alarm condition in real time; after the follow-up track ID is successfully identified, personnel information is obtained through key updating association, traditional intelligent analysis can only trigger the alarm condition and does not know who participates in, and the method can clearly know who participates in triggering the alarm condition without manual confirmation.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of artificial intelligence and security monitoring, and in particular to a police situation personnel dynamic association system and method based on multi-modal recognition. BACKGROUND

[0002] Limitations of traditional behavior recognition technology:

[0003] Dependence on image or video analysis: Traditional technology usually captures video images through cameras and then analyzes the actions in them using image processing algorithms.

[0004] Basic police situation type and location information: These technologies can identify abnormal behavior and issue alerts, but often only provide limited information such as the type and location of the police situation.

[0005] Application of deep learning in behavior recognition:

[0006] Improved recognition accuracy: In recent years, the development of deep learning technology has brought new breakthroughs in behavior recognition. For example, action recognition methods based on convolutional neural networks (CNN) can more accurately identify the identity and multiple actions of a person.

[0007] Multi-task processing capability: Deep learning models can handle multiple tasks such as identity recognition and action recognition simultaneously, providing more comprehensive behavior analysis.

[0008] However, current traditional behavior recognition technology (such as fight detection and crowd gathering detection) can only output basic police situation type and location information, but cannot associate the identity of the specific involved personnel. For example, when a fight occurs in the monitoring screen, the system can only indicate "fight detected", but cannot clearly identify the participants, resulting in low efficiency in subsequent handling. In addition, in existing technology, face recognition and behavior recognition are independent modules, lacking real-time data interaction mechanism, making it difficult to achieve accurate association in complex scenarios. SUMMARY

[0009] The purpose of the present application is to overcome the shortcomings of the prior art and propose a police situation personnel dynamic association system and method based on multi-modal recognition, which solves the problem of disconnection between behavior recognition and personnel identity through algorithm fusion and data binding.

[0010] The technical problem of the present application is solved by adopting the following technical solution:

[0011] The police situation personnel dynamic association system based on multi-modal recognition includes a face database module, an intelligent analysis server, an association control module, and a dynamic update module, wherein the face database module, the intelligent analysis server, the association control module, and the dynamic update module are connected in sequence.

[0012] Moreover, the face database module is used for storing face feature data of personnel and associated identity information.

[0013] Moreover, the intelligent analysis server is configured to receive video stream data, and synchronously run a face recognition algorithm and a behavior analysis algorithm.

[0014] Moreover, the association control module is used for matching a track ID generated by face recognition with alarm coordinates output by the behavior analysis algorithm, and binding personnel information and alarm events.

[0015] Moreover, the dynamic updating module is used for marking an alarm record by a unique key in the case that personnel are not identified when an alarm occurs, and updating associated information after subsequent successful identification.

[0016] An association method of a police alarm personnel dynamic association system based on multi-modal recognition, comprising the following steps:

[0017] Step 1: The face database module collects personnel face information and enters the face database.

[0018] Step 2: Configure an intelligent analysis task, add a camera and a channel in the device management module, add an intelligent analysis task by selecting a corresponding channel to be analyzed, click the algorithm rule configuration on the channel entry after completing the channel configuration, first configure a personnel recognition algorithm, and associate a face database required for comparison, and then configure other behavior analysis algorithms and check the personnel recognition to be associated.

[0019] Step 3: The intelligent analysis server pulls video stream, and synchronously performs face track tracking and behavior analysis.

[0020] Step 4: When the intelligent analysis server detects a behavior alarm, the association control module matches a face track ID in an alarm coordinate range, the face track ID also carries coordinates, and the coordinates of personnel in a fighting area are personnel participating in fighting, if the personnel recognition algorithm has identified the track ID as a specific person, the platform will carry personnel information when pushing an alarm, otherwise, the alarm carries a unique key, and the update of personnel information identified in step 5 is performed.

[0021] Step 5: If the track ID associated in the alarm area is not identified as a specific person when the alarm occurs, the dynamic updating module marks a unique key and pushes an alarm in real time; after subsequent successful identification of the track ID, the associated personnel information is updated through the key.

[0022] The advantages and positive effects of the present application are:

[0023] The application collects personnel face information through a face database module and enters the face database, configures an intelligent analysis task, adds a camera and a channel in a device management module, adds the intelligent analysis task by selecting a corresponding channel to be analyzed, an intelligent analysis server pulls a video stream, synchronously performs face track tracking and behavior analysis, and according to a detected behavior alarm, an association control module matches a face track ID in a coordinate range of the alarm, if the track ID associated in the alarm area is not recognized to be a specific personnel when the alarm occurs, a dynamic update module marks a unique key and pushes the alarm in real time; after a subsequent track ID is successfully recognized, personnel information is updated through the key, and a traditional intelligent analysis can only trigger an alarm and does not know who participates in the alarm, but the application can clearly know who participates in triggering the alarm and does not need to be confirmed by a person, for example, playing a mobile phone, and it is known who plays the mobile phone when the alarm occurs. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 The flowchart of the application. DETAILED DESCRIPTION

[0025] The application is further described below in combination with the drawings.

[0026] The alarm personnel dynamic association system based on multi-modal recognition includes a face database module, an intelligent analysis server, an association control module and a dynamic update module, wherein the face database module, the intelligent analysis server, the association control module and the dynamic update module are sequentially connected.

[0027] The face database module is used for storing face feature data of personnel and associated identity information. The intelligent analysis server is used for being configured to receive video stream data, synchronously running a face recognition algorithm and a behavior analysis algorithm. The association control module is used for matching a track ID generated by face recognition and alarm coordinates output by the behavior analysis algorithm, and binding personnel information and alarm events. The dynamic update module is used for marking an alarm record through a unique key in the case that personnel is not recognized when the alarm occurs, and updating associated information after subsequent successful recognition.

[0028] An association method of an alarm personnel dynamic association system based on multi-modal recognition, as shown in Figure 1 includes the following steps:

[0029] Step 1, the face database module collects personnel face information and enters the face database;

[0030] Step 2, configure an intelligent analysis task, add a camera and a channel in a device management module, add the intelligent analysis task by selecting a corresponding channel to be analyzed, configure an algorithm rule on an entry of the channel after the channel is configured, first configure a personnel recognition algorithm, associate a face database to be compared, and then configure other behavior analysis algorithms to check the personnel recognition to be associated;

[0031] Step 3, the intelligent analysis server pulls the video stream, and synchronously performs face track and behavior analysis (supplement: the video is synchronously analyzed by the personnel recognition algorithm and the specific behavior analysis algorithm such as fighting, the personnel recognition algorithm detects the personnel position to perform personnel tracking, and the behavior analysis algorithm searches whether there is a behavior violation behavior in the video) ;

[0032] Step 4, when the intelligent analysis server detects a behavior alarm (supplement: when the behavior alarm is detected, for example, fighting, the algorithm outputs the coordinates of the building area on the image), the association control module matches the face track ID in the alarm coordinate range, the face track ID also carries coordinates, and the personnel coordinates in the fighting area are the personnel participating in the fighting; if the personnel recognition algorithm has recognized who the track ID is at this time, the platform will carry the personnel information (for example, fighting, and what specific person participated in the fighting) when the alarm is pushed, and output the personnel information involved, otherwise, the alarm carries a unique key, and the personnel information is updated when the personnel information is recognized in the fifth step;

[0033] Step 5, if the track ID associated in the alarm area is not recognized when the alarm occurs, the dynamic update module marks the unique key and pushes the alarm in real time; after the track ID is successfully recognized subsequently, the personnel information is updated through the key, and the track ID algorithm will continuously try to recognize (supplement: if the alarm occurs)

[0034] It should be emphasized that the embodiments described in the present application are illustrative rather than limiting, and therefore the present application includes but is not limited to the embodiments described in the specific embodiments, and any other embodiments derived by those skilled in the art according to the technical solutions of the present application also belong to the scope of protection of the present application.

Claims

1. A police intelligence and personnel dynamic association system based on multimodal recognition, characterized by: It includes a face database module, an intelligent analysis server, an association control module and a dynamic update module, wherein the face database module, the intelligent analysis server, the association control module and the dynamic update module are connected in sequence.

2. The police information and personnel dynamic association system based on multimodal recognition according to claim 1 is characterized by: The face database module is used to store facial feature data of people and associated identity information.

3. The police information and personnel dynamic association system based on multimodal recognition according to claim 1 is characterized by: The intelligent analysis server is configured to receive video stream data and synchronously run a face recognition algorithm and a behavior analysis algorithm.

4. The police information and personnel dynamic association system based on multimodal recognition according to claim 1 is characterized by: The association control module is used to match the trajectory ID generated by face recognition with the police coordinates output by the behavior analysis algorithm, and bind personnel information with police events.

5. The police information and personnel dynamic association system based on multimodal recognition according to claim 1 is characterized by: The dynamic update module is used to mark the alarm record with a unique key when the person is not identified when the alarm occurs, and to update the associated information after subsequent successful identification.

6. A method for associating a police intelligence and personnel dynamic association system based on multimodal recognition according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: The face database module collects the facial information of the person and enters it into the face database; Step 2: Configure the intelligent analysis task. Add cameras and channels in the device management module. To add an intelligent analysis task, select the channel to be analyzed. After configuring the channel, click Algorithm Rule Configuration on the channel entry. First, configure the person recognition algorithm and associate it with the face database that needs to be compared. Then configure other behavior analysis algorithms and check the ones to be associated with person recognition. Step 3: The intelligent analysis server pulls the video stream and performs face tracking and behavior analysis simultaneously; Step 4. When the intelligent analysis server detects a behavioral alarm, the association control module matches the face track ID within the alarm coordinate range. The face track ID also carries coordinates. The coordinates of the people in the fighting area are the people involved in the fight. If the person recognition algorithm has identified the owner of this track ID at this time, the platform will output the information of the person involved with the person information when pushing the alarm. Otherwise, this alarm carries the alarm unique key and updates the person information after step 5. Step 5: If the associated trajectory ID in the alarm area does not identify a specific person when an alarm occurs, the dynamic update module will mark the unique key and push the alarm in real time; after the subsequent trajectory ID is successfully identified, the associated person information is updated through the key.

Images