Intelligent visual perception distributed node cooperation method, device and system

Abstract

The application discloses a kind of intelligent visual perception distributed node cooperation method, device and system, purpose is to solve the network delay, congestion and other problems when multi-camera cross mirror linkage scene, wherein, method includes: obedient node continuously obtains at least one camera transmission video information governed by it;Built-in convolutional neural network algorithm is scheduled from video information to identify preset target and judge whether preset target appears interesting event, if yes, then interesting information and data file are stored on obedient node;Continuously listen to the control instruction information sent by control node in node cluster;Based on control instruction information, the interesting information, data file stored on the obedient node are transmitted to control node.The application improves the response speed of cross mirror tracking, search and other by deploying the computing power and algorithm of artificial intelligence analysis near camera cluster.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence. Specifically, the present invention relates to an intelligent visual perception distributed node cooperation method, device and system. Background Art

[0002] Artificial intelligence analysis using technologies such as convolutional neural networks has been applied in fields such as community security and mine safety production. The known system uses a high-performance central server to perform video analysis. The central centralized visual analysis method requires all video streams to be centrally pushed to the central server for video stream analysis, which has great disadvantages for scenarios with a large number of cameras in communities, banks, coal mines, etc., and scenarios involving cross-lens search and tracking of personnel. In addition, simultaneous access of multiple video sources is likely to cause network congestion, and there is a large delay for some video sources to reach the central server for video stream analysis, making it difficult to meet the requirements of concurrent real-time responses such as cross-lens tracking. Summary of the Invention

[0003] In order to solve problems such as network delay and congestion in the scenario of cross-lens linkage of multiple cameras, the present invention proposes an intelligent visual perception distributed node cooperation method, device and system.

[0004] Solution 1:

[0005] Provide an intelligent visual perception distributed node cooperation method, which includes the following steps:

[0006] The compliant node continuously obtains video information transmitted by at least one camera under its jurisdiction;

[0007] Dispatch the built-in convolutional neural network algorithm to identify a preset target from the video information and determine whether an interesting event occurs for the preset target. If so, store the interesting information corresponding to the interesting event and the data file output by the convolutional neural network algorithm on the compliant node;

[0008] Continuously listen for control instruction information sent by the control node in the node cluster, where the node cluster includes at least one compliant node and a control node communicatively connected to the compliant node;

[0009] Transmit the interesting information and data file stored on the compliant node to the control node based on the control instruction information.

[0010] Preferably, the compliant node is an intelligent visual perception node. Before the compliant node obtains video information transmitted by at least one camera under its jurisdiction, it includes the following steps:

[0011] At least one camera is grouped according to a preset grouping rule, and a compliant node is configured for each group of cameras; the compliant node communicates with the at least one camera via a video stream protocol.

[0012] Preferably, determining whether the preset target has experienced an event of interest includes the following steps:

[0013] The system determines whether an event of interest has occurred at the preset target based on predefined rules, which are either control command information sent by the control node or custom rules.

[0014] Preferably, the steps before storing the interest information corresponding to the event of interest and the data file output by the convolutional neural network algorithm on the conformity node include the following:

[0015] Determine whether the information of interest meets the predefined conditions. If so, store the information of interest corresponding to the event of interest and the data file output by the convolutional neural network algorithm on the conformity node.

[0016] Preferably, before transmitting the information of interest and data file stored on the compliant node to the controlling node based on the control command information, the following steps are included:

[0017] The information of interest corresponding to the event of interest and the data file output by the convolutional neural network algorithm are uploaded to the central storage server for backup.

[0018] Option 2:

[0019] A method for intelligent visual perception distributed node collaboration is provided, the method comprising the following steps:

[0020] The control node acquires control command information and sends the control command information to the compliant node;

[0021] The system receives the information of interest and the data file from all compliant nodes in the receiving node cluster that are connected to the control node, based on the control command information.

[0022] The information of interest and the data file are merged according to preset rules and then fed back.

[0023] Preferably, before fusing the information of interest and the data file according to preset rules and feeding it back, the following steps are included:

[0024] Listen to the responses of all compliant nodes and determine whether the response time of the compliant nodes exceeds the preset response time threshold.

[0025] If so, the corresponding information of interest and the data file are retrieved from the central storage server.

[0026] Option 3:

[0027] A smart visual perception distributed node collaboration device is provided, the smart visual perception distributed node collaboration device comprising:

[0028] The video acquisition module is used by compliant nodes to continuously acquire video information transmitted from at least one camera under their jurisdiction.

[0029] The scheduling and judgment module is used to schedule the built-in convolutional neural network algorithm to identify a preset target from the video information and determine whether the preset target has an event of interest. If so, the interest information corresponding to the event of interest and the data file output by the convolutional neural network algorithm are stored on the obedient node.

[0030] The monitoring and control module is used to continuously monitor control command information sent by the control node in the node cluster, wherein the node cluster includes at least one subordinating node and a control node that is communicatively connected to the subordinating node.

[0031] The storage module is used to transmit the information of interest and data files stored on the compliant node to the control node based on the control command information.

[0032] Option 4:

[0033] A smart visual perception distributed node collaboration device is provided, the smart visual perception distributed node collaboration device comprising:

[0034] The acquisition and transmission module is used for the control node to acquire control command information and send the control command information to the compliant node;

[0035] The receiving module is used to receive the information of interest and the data file from all obedient nodes in the node cluster that are connected to the control node, based on the control command information.

[0036] The fusion feedback module is used to fuse the information of interest and the data file according to preset rules and then provide feedback.

[0037] Option 5:

[0038] A smart visual perception distributed node collaboration system is provided. The smart visual perception distributed node collaboration system includes a central storage server and at least one node cluster connected to the central storage server. The node cluster includes at least one subordinating node and a control node communicatively connected to the subordinating node.

[0039] The subservient node continuously acquires video information transmitted from at least one camera under its jurisdiction; it schedules a built-in convolutional neural network algorithm to identify a preset target from the video information and determines whether the preset target exhibits an event of interest. If so, it stores the interest information corresponding to the event of interest and the data file output by the convolutional neural network algorithm on the subservient node; it continuously monitors control command information sent by the control node in the node cluster; and based on the control command information, it transmits the interest information and data file stored on the subservient node to the control node.

[0040] The central storage server is used to back up the interest information corresponding to the interest event reported by the compliant node and the data file output by the convolutional neural network algorithm, and to transmit the backed-up interest information and the data file to the control node;

[0041] The controlling node acquires control command information and sends the control command information to the compliant node; it receives the information of interest and the data file from all compliant nodes connected to the controlling node in the node cluster based on the control command information; it merges the information of interest and the data file according to a preset rule and feeds it back.

[0042] Compared with existing technologies, the intelligent visual perception distributed node collaboration method, device, and system of the present invention have the following beneficial effects:

[0043] This invention discloses an intelligent visual perception distributed node collaboration method, device, and system. By deploying the computing power and algorithms of artificial intelligence analysis near the camera, it shortens the reflection arc, improves the response speed, avoids network latency and congestion, reduces the overall network bandwidth resource consumption of the system, and achieves fast response for high-concurrency searches.

[0044] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of the invention. Attached Figure Description

[0045] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0046] Figure 1 This is a schematic diagram of an application scenario and flowchart of an intelligent visual perception distributed node collaboration method according to an embodiment of the present invention.

[0047] Figure 2 This is a flowchart illustrating an intelligent visual perception distributed node collaboration method according to an embodiment of the present invention.

[0048] Figure 3 This is another flowchart illustrating an intelligent visual perception distributed node collaboration method according to an embodiment of the present invention;

[0049] Figure 4 This is a schematic diagram of the module structure of an intelligent visual perception distributed node collaboration device according to another embodiment of the present invention;

[0050] Figure 5 This is a schematic diagram of another module structure of an intelligent visual perception distributed node collaboration device according to another embodiment of the present invention. Detailed Implementation

[0051] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0052] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 202, 204, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Please see Figure 1 This invention provides an intelligent visual perception distributed node collaboration method, applied to cross-camera linkage based on multiple cameras, such as cross-camera search and tracking of people. For example,... Figure 1The cross-camera tracking system shown includes a central storage server 800 and a node cluster 8 connected to the central storage server 800. The node cluster 8 includes a control node 88, a first subservient node 881, a second subservient node 883, a first camera 8811, a second camera 8813, a third camera 8831, and a fourth camera 8833. The first camera 8811 and the second camera 8813 are respectively connected to the first subservient node 881, the third camera 8831 and the fourth camera 8833 are respectively connected to the second subservient node 883, and the first subservient node 881 and the second subservient node 883 are respectively connected to the control node 88. In the specific implementation process, the first camera 8811, the second camera 8813, the third camera 8831, and the fourth camera 8833 are distributed in different locations. The first camera 8811, the second camera 8813, and the first subservient node 881 are configured in the same local area network. The third camera 8831, the fourth camera 8833, and the second subservient node 883 are also configured in the same local area network. The control node 88 is preferentially configured on the wide area network. The control node 88, the first subservient node 881, and the second subservient node 883 are connected to the central storage server 800 through the network.

[0055] Please see Figure 2 , Figure 2 This diagram illustrates a flowchart of a distributed node collaboration method for intelligent visual perception according to an embodiment of the present invention. For ease of understanding of this embodiment, it is now presented as follows. Figure 2 The intelligent visual perception distributed node collaboration method shown is applied to the Figure 1 The cross-camera tracking system shown is used as an example for explanation and illustration. It should be understood that this explanation and illustration should not be construed as a limitation on the intelligent visual perception distributed node cooperation method of the present invention. Specifically, the intelligent visual perception distributed node cooperation method of the present invention, applied to compliant nodes, includes the following steps:

[0056] Step S101: The compliant node continuously acquires video information transmitted from at least one camera under its jurisdiction. The compliant node is an intelligent visual perception node, comprising an embedded processor, embedded dynamic random access memory, a mobile low-power graphics processing unit, dedicated neural network inference hardware, dedicated video encoding / decoding hardware, and also includes non-volatile storage devices and a power management module. The embedded processor, mobile low-power graphics processing unit, and dedicated neural network inference hardware provide the heterogeneous computing power required for artificial intelligence algorithms. The embedded dynamic random access memory is a cache pool for data processing; the non-volatile memory is used for permanent data storage; and the power management module supplies power to the compliant node and adjusts the power supply to the embedded processor according to its load.

[0057] In some implementations, before a compliant node acquires video information transmitted from at least one camera under its jurisdiction, the following steps are included:

[0058] At least one camera is grouped according to a preset grouping rule, and a compliant node is configured for each group of cameras; the compliant node communicates with at least one camera via a video stream protocol.

[0059] For example, such as Figure 1 The cross-camera tracking system shown has the first camera 8811, the second camera 8813, the third camera 8831, and the fourth camera 8833 grouped in pairs, respectively assigned to the first subservient node 881 and the second subservient node 883. It is worth noting that in reality, the number of groups may be more than two, such as five or twenty. This embodiment of the invention does not impose such a limitation, and the preset grouping rules can be defined by the implementer according to the application scenario.

[0060] Preferably, the video streaming protocol is the Real-Time Streaming Protocol (RTSP), an application layer protocol in the TCP / IP protocol suite, and an IETF RFC standard submitted by Columbia University, Netscape, and RealNetworks. This protocol defines how one-to-many applications can efficiently transmit multimedia data over IP networks.

[0061] Step S103: The built-in convolutional neural network algorithm is scheduled to identify a preset target from the video information and determine whether the preset target has an event of interest. If so, the interest information corresponding to the event of interest and the data file output by the convolutional neural network algorithm are stored on the obedient node.

[0062] In some implementations, the convolutional neural network algorithm is a deep learning object detection method based on regression methods, such as the YOLO neural network and the SSD neural network. YOLO is an object recognition and localization algorithm based on deep neural networks. Its biggest feature is that it runs very fast and can be used in real-time systems. The core of the SSD neural network is to use small convolutional filters to predict the class scores and position offsets of a fixed set of default bounding boxes on the feature map.

[0063] In some implementations, the preset target is preferably a person, object, or the like.

[0064] In some implementations, determining whether an event of interest has occurred at a preset target includes the following steps:

[0065] The system determines whether an event of interest has occurred at a preset target based on predefined rules, which are either control commands sent by the control node or custom rules.

[0066] For example, when a control node sends an output displaying, for instance, all the paths and video clips a fugitive has taken over the past fourteen days, the fugitive's escape route is considered an event of interest. Furthermore, events such as people falling or vehicles crossing boundaries can also be defined as events of interest; therefore, this embodiment of the invention does not impose any limitations on this.

[0067] Preferably, the steps before storing the data files corresponding to the events of interest and the output of the convolutional neural network algorithm on the conformity-type nodes include:

[0068] Determine whether the information of interest meets the predefined conditions. If so, store the information of interest corresponding to the event of interest and the data file output by the convolutional neural network algorithm on the obedient node.

[0069] In some implementations, predefined conditions are based on parameter confirmation of convolutional neural network algorithms, i.e., conditional interventions are performed during the process of target recognition, target classification, target comparison, target tracking, etc., of video information.

[0070] Step S105: Continuously monitor the control command information sent by the control node in the node cluster, wherein the node cluster includes at least one of the subservient nodes and a control node that is communicatively connected to the subservient node.

[0071] In some implementations, the compliant node continuously receives video images from the cameras under its jurisdiction and schedules the convolutional neural network algorithm built into the compliant node to identify targets such as people and objects in the images. Based on predefined rules, it determines whether an event of interest has occurred, such as a person falling or a vehicle crossing the boundary. If an event of interest is found, the node saves the information of interest, such as the image corresponding to the event of interest, and the data file output by the convolutional neural network algorithm, to the local storage of the compliant node. At the same time, the relevant information is transmitted to the central storage server for long-term data backup through a message broker.

[0072] Preferably, when configuring the local storage capacity of the compliant node, it can support approximately 14 to 28 days of storage; when the local storage capacity is exhausted, the oldest historical data is cyclically overwritten.

[0073] Step S107: Based on the control command information, transmit the information of interest and data file stored on the compliant node to the control node.

[0074] In some implementations, the information of interest includes images, data files, and video clips. Before transmitting the information of interest and data files stored on the compliant node to the control node based on control command information, image comparison technology is used to search for all images that reach a similarity threshold in the memory of the compliant node according to the control command information, and the relevant images, data files, and video clips are transmitted to the location specified by the control node.

[0075] For example, when a compliant node receives a fugitive search instruction from a controlling node, it performs a secondary analysis and comparison of historical personnel images stored locally, and then uploads video clips to the controlling node.

[0076] In some implementations, before transmitting the information of interest and data files stored on the compliant node to the controlling node based on control command information, the following steps are included:

[0077] The data files containing the information of interest corresponding to the events of interest and the output of the convolutional neural network algorithm are uploaded to the central storage server for backup.

[0078] Please see Figure 3 Based on the same inventive concept as the intelligent visual perception distributed node collaboration method in the embodiments of the present invention, the present invention also provides an intelligent visual perception distributed node collaboration method, which is applied to control nodes and includes the following steps:

[0079] Step T202: The control node acquires control command information and sends it to the compliant node. Specifically, unlike the compliant node, the control node supports human-computer interaction and receives commands from the user, such as displaying all paths and video clips spliced ​​together by a fugitive over the past fourteen days. The control node can, for example, divide a cross-border fugitive search request into multiple sub-commands based on the group to which the involved cameras belong, and send each sub-command to the compliant node corresponding to that group.

[0080] In some implementations, the controlling node and the subservient node communicate through a message middleware. The subservient node periodically reports its own status to the controlling node and continuously listens for relevant control command information from the controlling node.

[0081] Step T204: Receive the information of interest and the data file from all compliant nodes in the node cluster that are connected to the control node, based on the control command information.

[0082] Step T206: Merge the information of interest and the data file according to preset rules and provide feedback.

[0083] Specifically, the control node merges, sorts, and integrates information of interest at a specified location according to its relevance such as time and location, and pushes the fusion result to the user in a manner suitable for human-computer interaction.

[0084] In some implementations, in order to reduce the resource consumption of the control node, the compliant node can also search and merge the results of the multiple cameras under its jurisdiction, and then send the merged results to the control node, which will then fuse all the received results and push the fused results to the user in a manner suitable for human-computer interaction.

[0085] In some implementations, after a compliant node receives control command information from a controlling node, it first replies to the controlling node to confirm receipt. Then, according to the control command information, it transmits the information of interest and data files stored on the compliant node to the controlling node. It is worth noting that there are multiple compliant nodes. Therefore, the node cluster in this embodiment is not simply the sum of the number of nodes, but possesses superpowers different from a single node, featuring low power consumption, high computing power, and support for high-concurrency searches.

[0086] In some implementations, to prevent some compliant nodes from failing, the control node listens for feedback from all compliant nodes. If some failed compliant nodes fail to provide feedback within a timeout period, the control node initiates its own visual perception backup service, sending the search task for the failed node to the backup service. The backup service then searches the central storage server for relevant information of interest previously uploaded by the corresponding node and transmits the search results to the designated location of the control node. Before fusing the information of interest and data files according to preset rules and providing feedback, the process includes the following steps:

[0087] Listen to the responses of all compliant nodes and determine whether the response time of the compliant nodes exceeds the preset response time threshold.

[0088] If so, the corresponding information and data files of interest are retrieved from the central storage server.

[0089] In some implementations, the camera supports motion detection. When it detects people or objects in the captured image, it automatically saves the video clip and sends the video clip to the compliant node. By leveraging the advantages of weak intelligence, such as motion detection, personnel detection, and boundary detection, and the complex algorithms of edge compliant nodes, the integration of control node scheduling with multiple compliant nodes effectively improves the real-time response characteristics of the system and further reduces network overhead.

[0090] Please see Figure 4 Based on the same inventive concept as the intelligent visual perception distributed node collaboration method in an embodiment of the present invention, another embodiment of the present invention provides an intelligent visual perception distributed node collaboration device, which is applied to compliant nodes, including:

[0091] The video acquisition module 4001 is used for compliant nodes to continuously acquire video information transmitted by at least one camera under their jurisdiction;

[0092] The scheduling and judgment module 4003 is used to schedule the built-in convolutional neural network algorithm to identify the preset target from the video information and determine whether the preset target has an event of interest. If so, the interest information corresponding to the event of interest and the data file output by the convolutional neural network algorithm are stored on the obedient node.

[0093] The monitoring and control module 4005 is used to continuously monitor the control command information sent by the control node in the node cluster, wherein the node cluster includes at least one subordinating node and a control node that is communicatively connected to the subordinating node.

[0094] Storage module 4007 is used to transmit information of interest and data files stored on the compliant node to the control node based on control command information.

[0095] Preferably, the compliant node is an intelligent visual perception node, and the intelligent visual perception distributed node collaboration device includes:

[0096] The group configuration module is used to group at least one camera according to a preset grouping rule and configure a corresponding subordinating node for each group of cameras before the subordinating node obtains the video information transmitted by at least one camera under its jurisdiction; the subordinating node and at least one camera communicate and connect through a video stream protocol.

[0097] In some implementations, the scheduling determination module 4003 includes:

[0098] The predefined unit is used to determine whether a preset target has an event of interest based on predefined rules. The predefined rules are based on control command information sent by the control node or are custom-defined.

[0099] Preferably, the intelligent visual perception distributed node collaboration device further includes:

[0100] The judgment unit is used to determine whether the information of interest meets the predefined conditions before storing the information of interest corresponding to the event of interest and the data file output by the convolutional neural network algorithm on the subordinating node. If so, the information of interest corresponding to the event of interest and the data file output by the convolutional neural network algorithm are stored on the subordinating node.

[0101] Preferably, the intelligent visual perception distributed node collaboration device further includes:

[0102] The upload backup module is used to upload the information of interest corresponding to the event of interest and the data files output by the convolutional neural network algorithm to the central storage server for backup before transmitting the information of interest and data files stored on the compliant node to the control node based on the control command information.

[0103] Please see Figure 5 Correspondingly, based on the same inventive concept as the intelligent visual perception distributed node collaboration device provided in another embodiment of the present invention, another embodiment of the present invention provides an intelligent visual perception distributed node collaboration device, which is applied to a control node, including:

[0104] The acquisition and transmission module 5002 is used by the control node to acquire control command information and send the control command information to the compliant node.

[0105] The receiving module 5004 is used to receive the information of interest and the data file from all obedient nodes connected to the control node in the node cluster, based on the control command information.

[0106] The fusion feedback module 5006 is used to fuse information of interest and data files according to preset rules and then provide feedback.

[0107] In some embodiments, the intelligent visual perception distributed node collaboration device further includes:

[0108] The monitoring and feedback module is used to monitor the responses of all obedient nodes before fusing information of interest and data files according to preset rules and providing feedback, and to determine whether the response time of obedient nodes exceeds the preset response time threshold.

[0109] The central acquisition module is used to retrieve the corresponding information of interest and data files from the central storage server when the response time of the compliant node exceeds a preset response time threshold.

[0110] Based on the same inventive concept as the intelligent visual perception distributed node collaboration method of the present invention and the intelligent visual perception distributed node collaboration device of another embodiment of the present invention, another embodiment of the present invention provides an intelligent visual perception distributed node collaboration system, which includes a central storage server and at least one node cluster connected to the central storage server. The node cluster includes at least one subordinating node and a control node that is communicatively connected to the subordinating node.

[0111] The subservient node continuously acquires video information transmitted from at least one camera under its jurisdiction; it schedules the built-in convolutional neural network algorithm to identify a preset target from the video information and determines whether an event of interest has occurred on the preset target. If so, it stores the information of interest corresponding to the event of interest and the data file output by the convolutional neural network algorithm on the subservient node; it continuously listens for control command information sent by the control node in the node cluster; and based on the control command information, it transmits the information of interest and the data file stored on the subservient node to the control node.

[0112] The central storage server is used to back up the information of interest corresponding to the events of interest reported by the compliant nodes and the data files output by the convolutional neural network algorithm, as well as to transmit the backed-up information of interest and data files to the control nodes;

[0113] The control node acquires control command information and sends it to the compliant node; it receives information of interest and the data file from all compliant nodes connected to the control node in the node cluster based on the control command information; it merges the information of interest and the data file according to a preset rule and feeds it back.

[0114] It is worth noting that another embodiment of the present invention provides an intelligent visual perception distributed node collaboration system. The specific execution steps and principles are the same as those of the two embodiments described above, so they will not be repeated here.

[0115] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0116] Compared with existing technologies, the intelligent visual perception distributed node collaboration method, device, and system of the present invention have the following beneficial effects:

[0117] This invention discloses an intelligent visual perception distributed node collaboration method, device, and system. By deploying the computing power and algorithms of artificial intelligence analysis in a node cluster near cameras, the reflection arc is shortened, the response speed is improved, network latency and congestion are avoided, the overall network bandwidth resource consumption of the system is reduced, and a fast response to high-concurrency search is achieved.

[0118] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0119] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0120] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0121] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0122] The above description is only a partial embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent visual perception distributed node cooperation, characterized in that, The intelligent visual perception distributed node collaboration method includes the following steps: The compliant node continuously acquires video information transmitted by at least one camera under its jurisdiction; the compliant node is an intelligent visual perception node deployed near the camera, including an embedded processor, dedicated neural network inference hardware, dedicated video encoding and decoding hardware and non-volatile storage hardware, with heterogeneous computing power for artificial intelligence algorithms, and a built-in convolutional neural network algorithm. The convolutional neural network algorithm built into the obedient node is scheduled to identify a preset target from the video information and determine whether the preset target has an event of interest. If so, the interest information corresponding to the event of interest and the data file output by the convolutional neural network algorithm are stored on the obedient node. The submissive node continuously listens for control command information sent by the control node in the node cluster, wherein the node cluster includes at least one submissive node and a control node that is communicatively connected to the submissive node. The interest information corresponding to the event of interest and the data file output by the convolutional neural network algorithm are uploaded to the central storage server for backup. Based on the control command information, the information of interest and data files stored on the compliant node are transmitted to the control node. 2.The intelligent visual perception distributed node cooperation method of claim 1, wherein, The submissive node is an intelligent visual perception node. Before acquiring video information transmitted by at least one camera under its jurisdiction, the submissive node includes the following steps: At least one camera is grouped according to a preset grouping rule, and a compliant node is configured for each group of cameras; the compliant node communicates with the at least one camera via a video stream protocol.

3. The intelligent visual perception distributed node collaboration method as described in claim 1, characterized in that, The step of determining whether the preset target has encountered an event of interest includes the following steps: The system determines whether an event of interest has occurred at the preset target based on predefined rules, which are either control command information sent by the control node or custom rules.

4. The intelligent visual perception distributed node collaboration method as described in claim 1, characterized in that, Before storing the interest information corresponding to the interest event and the data file output by the convolutional neural network algorithm on the conformity node, the following steps are included: Determine whether the information of interest meets the predefined conditions. If so, store the information of interest corresponding to the event of interest and the data file output by the convolutional neural network algorithm on the conformity node.

5. A method for intelligent visual perception distributed node collaboration, characterized in that, The intelligent visual perception distributed node collaboration method includes the following steps: The control node acquires control command information and sends it to the compliant node, which is an intelligent visual perception node deployed near the camera. This node includes an embedded processor, dedicated neural network inference hardware, dedicated video encoding / decoding hardware, and non-volatile storage hardware. It possesses heterogeneous computing power for artificial intelligence algorithms and has a built-in convolutional neural network algorithm. The control node monitors the responses of all compliant nodes and determines whether the response time of each node exceeds a preset response time threshold. If so, it retrieves the corresponding information of interest and data files from the central storage server. The system receives the information of interest and the data file from all compliant nodes in the receiving node cluster that are connected to the control node, based on the control command information. The information of interest and the data file are merged according to preset rules and then fed back.

6. A smart visual perception distributed node collaboration device, characterized in that, This intelligent visual perception distributed node collaboration device includes: The video acquisition module is used for the compliant node to continuously acquire video information transmitted by at least one camera under its jurisdiction. The compliant node is an intelligent visual perception node deployed near the camera, including an embedded processor, dedicated neural network inference hardware, dedicated video encoding and decoding hardware and non-volatile storage hardware, with heterogeneous computing power for artificial intelligence algorithms and built-in convolutional neural network algorithms. The scheduling and judgment module is used to schedule the convolutional neural network algorithm built into the obedient node to identify a preset target from the video information and determine whether the preset target has an event of interest. If so, the interest information corresponding to the event of interest and the data file output by the convolutional neural network algorithm are stored on the obedient node. The monitoring and control module is used for the compliant node to continuously monitor the control command information sent by the control node in the node cluster, wherein the node cluster includes at least one compliant node and a control node that is communicatively connected to the compliant node. The storage module is used to transmit the information of interest and data files stored on the compliant node to the control node based on the control command information.

7. An intelligent visual perception distributed node collaboration system, characterized in that, The intelligent visual perception distributed node collaboration system includes a central storage server and at least one node cluster connected to the central storage server. The node cluster includes at least one subservient node and a control node communicatively connected to the subservient node. The subservient node is an intelligent visual perception node deployed near a camera, including an embedded processor, dedicated neural network inference hardware, dedicated video encoding and decoding hardware, and non-volatile storage hardware. It has heterogeneous computing power for artificial intelligence algorithms and a built-in convolutional neural network algorithm. The subservient node continuously acquires video information transmitted from at least one camera under its jurisdiction. It schedules the built-in convolutional neural network algorithm to identify a preset target from the video information and determine whether the preset target has an event of interest. If so, it stores the interest information corresponding to the event of interest and the data file output by the convolutional neural network algorithm on the subservient node. The system continuously monitors control command information sent by the control node in the node cluster; based on the control command information, it transmits the information of interest and data files stored on the compliant node to the control node. The central storage server is used to back up the interest information corresponding to the interest event reported by the compliant node and the data file output by the convolutional neural network algorithm, and to transmit the backed-up interest information and the data file to the control node; The control node acquires control command information and sends the control command information to the compliant node; The system receives the information of interest and the data file from all compliant nodes in the receiving node cluster that are connected to the control node, based on the control command information. The information of interest and the data file are merged according to preset rules and then fed back.

Images