Determination method of situation recognition model, object situation recognition method and related equipment

Abstract

The embodiment of the invention provides a situation recognition model determination method, an object situation recognition method and related equipment, and the method is applied to a base station and specifically comprises the steps: receiving a first model parameter sent by an edge node; obtaining a plurality of samples and a sample label of each sample; training a situation recognition model by using the plurality of samples and the sample label of each sample to obtain a trained situation recognition model; the model parameter variable quantity is sent to the edge node; and receiving the updated first model parameter sent by the edge node, and returning to execute the step of obtaining the plurality of samples and the sample label of each sample until the trained situation recognition model meets a preset condition, thereby obtaining a target situation recognition model. According to the embodiment of the invention, the accuracy of situation recognition is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of smart communities, and in particular, relates to a method for determining a situation recognition model, an object situation recognition method, and related equipment. Background Art

[0002] Situational awareness in a smart community is the ability to gain insight into community security risks in a dynamic and holistic manner based on the environment. It uses technologies such as the Internet of Things, cloud computing, big data, and artificial intelligence to conduct real-time monitoring, analysis, and early warning of various resources within the community in order to improve the community's security management level and emergency response capabilities.

[0003] Existing situational awareness methods generally deploy a large number of high-definition cameras, facial recognition equipment, environmental monitoring sensors, etc. in smart communities to conduct real-time monitoring of the flow of people in the community, environmental changes when vehicles enter, etc., in order to achieve situational awareness in smart communities. However, the accuracy of situational awareness of this method is relatively low. Summary of the invention

[0004] The embodiments of the present application provide a method for determining a situation recognition model, an object situation recognition method, and related equipment, which solve the problem of low accuracy of situation perception in real life.

[0005] In a first aspect, an embodiment of the present application provides a method for determining a situation recognition model, which is applied to a base station, and the method includes:

[0006] Receiving a first model parameter sent by an edge node;

[0007] Acquire multiple samples and a sample label of each sample, each sample including communication perception information of a first object; the sample label of each sample is used to characterize the object state of the first object;

[0008] Using multiple samples and sample labels of each sample to train a situation recognition model, the situation recognition model is constructed based on the first model parameters to obtain a trained situation recognition model, and the trained situation recognition model includes the second model parameters;

[0009] Sending a model parameter change amount to the edge node, where the model parameter change amount is a difference between the second model parameter and the first model parameter;

[0010] Receive the updated first model parameter sent by the edge node, where the updated first model parameter is determined by the edge node based on the model parameter change sent by multiple base stations, and return to execute the step of obtaining multiple samples and the sample label of each sample until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model.

[0011] In a second aspect, a method for determining a situation recognition model is provided, which is applied to an edge node, and the method includes:

[0012] Sending first model parameters to multiple base stations respectively, so that each base station obtains multiple samples and a sample label of each sample, each sample includes communication perception information of the first object; the sample label of each sample is used to characterize the object situation of the first object, and using the multiple samples and the sample label of each sample to train a situation recognition model, the situation recognition model is constructed based on the first model parameters, and a trained situation recognition model is obtained, and the trained situation recognition model includes second model parameters;

[0013] receiving model parameter changes respectively sent by a plurality of base stations, where the model parameter changes are differences between the second model parameters and the first model parameters;

[0014] The updated first model parameters are sent to multiple base stations respectively, and the updated first model parameters are determined by the edge node based on the model parameter changes sent by the multiple base stations, so that each base station returns to execute the step of obtaining multiple samples and the sample label of each sample until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model.

[0015] In a third aspect, a method for object situation recognition is provided, which is applied to a base station, and the method includes:

[0016] Acquire communication perception information of the object to be identified;

[0017] The target situation recognition model determined by the situation recognition model determination method provided by any optional implementation manner in the first aspect or the second aspect is used to process the communication perception information of the object to be recognized to obtain the object situation of the object to be recognized.

[0018] In a fourth aspect, an embodiment of the present application provides a device for determining a situation recognition model, which is applied to a base station, and the device includes:

[0019] A receiving module, used for receiving a first model parameter sent by an edge node;

[0020] An acquisition module, used to acquire multiple samples and a sample label of each sample, each sample including communication perception information of the first object; the sample label of each sample is used to characterize the object state of the first object;

[0021] A training module is used to train a situation recognition model using multiple samples and a sample label of each sample, wherein the situation recognition model is constructed based on the first model parameter to obtain a trained situation recognition model, wherein the trained situation recognition model includes the second model parameter;

[0022] A sending module, used for sending a model parameter change amount to an edge node, where the model parameter change amount is a difference between the second model parameter and the first model parameter;

[0023] The receiving module is also used to receive the updated first model parameters sent by the edge node, where the updated first model parameters are determined by the edge node based on the model parameter changes sent by multiple base stations, and return to execute the steps of obtaining multiple samples and the sample label of each sample until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model.

[0024] In a fifth aspect, an embodiment of the present application provides a device for determining a situation recognition model, which is applied to an edge node, and the device includes:

[0025] A sending module, used to send the first model parameters to the multiple base stations respectively, so that each base station obtains multiple samples and a sample label of each sample, each sample includes communication perception information of the first object; the sample label of each sample is used to characterize the object situation of the first object, and the situation recognition model is trained using the multiple samples and the sample label of each sample, the situation recognition model is constructed based on the first model parameters, and a trained situation recognition model is obtained, and the trained situation recognition model includes a second model parameter;

[0026] A receiving module, used for receiving model parameter changes respectively sent by a plurality of base stations, where the model parameter change is a difference between the second model parameter and the first model parameter;

[0027] The sending module is also used to send the updated first model parameters to multiple base stations respectively. The updated first model parameters are determined by the edge node based on the model parameter changes sent by the multiple base stations, so that each base station returns to execute the steps of obtaining multiple samples and the sample label of each sample until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model.

[0028] In a sixth aspect, an embodiment of the present application provides an object situation recognition device, which is applied to a base station, and the device includes:

[0029] An acquisition module, used to acquire communication perception information of an object to be identified;

[0030] A processing module is used to process the communication perception information of the object to be identified using the target situation recognition model determined by the situation recognition model determination method provided by any optional implementation method of the first aspect or the second aspect to obtain the object situation of the object to be identified.

[0031] In the seventh aspect, an electronic device is provided, comprising: a memory for storing computer program instructions; a processor for reading and running the computer program instructions stored in the memory to execute the method for determining the situation recognition model provided by any optional implementation method in the first aspect or the second aspect, or to execute the object situation recognition method provided by the third aspect.

[0032] In an eighth aspect, a computer storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method for determining a situation recognition model provided by any optional implementation method in the first aspect or the second aspect is implemented, or the object situation recognition method provided by the third aspect is implemented.

[0033] In the ninth aspect, a computer program product is provided, which includes a computer program. When the computer program is executed by a processor, it implements the method for determining the situation recognition model provided by any optional implementation method in the first aspect or the second aspect, or implements the object situation recognition method provided by the third aspect.

[0034] In an embodiment of the present application, the edge node can send the first model parameter to multiple base stations respectively, so that each base station can obtain multiple samples and the sample label of each sample, wherein each sample includes the communication perception information of the first object, and the sample label of each sample is used to characterize the object situation of the first object, and the situation recognition model constructed based on the first model parameter is trained using the multiple samples and the sample label of each sample to obtain the trained situation recognition model, and the trained situation recognition model includes the second model parameter. In this way, the edge node can receive the model parameter change amount sent by multiple base stations respectively, and the model parameter change amount is the difference between the second model parameter and the first model parameter, that is, it characterizes the change before and after the training of the above situation recognition model, and sends the updated first model parameter to multiple base stations respectively, so that each base station continues to repeatedly train the situation recognition model constructed based on the first model parameter until the trained situation recognition model meets the preset conditions, and the target situation recognition model is obtained. Since the above-mentioned updated first model parameters are determined by the edge node based on the model parameter changes sent by multiple base stations, that is, after the base station completes training of the situation recognition model each time, the edge node can determine the model parameters used to reconstruct the situation recognition model based on the changes before and after the training of the situation recognition model of each base station, so that each base station continues to train the situation recognition model until the trained situation recognition model meets the preset conditions. In this way, a more accurate target situation recognition model can be obtained, and then in the subsequent situation recognition process, the accuracy of situation recognition can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0036] Figure 1 It is an architecture diagram between a visual perception device, a base station, and an edge node provided in an embodiment of the present application;

[0037] Figure 2 This is one of the flow charts of a method for determining a situation recognition model provided in an embodiment of the present application;

[0038] Figure 3 This is a second flow chart of a method for determining a situation recognition model provided in an embodiment of the present application;

[0039] Figure 4 This is a flowchart of a method for determining a situation recognition model provided in an embodiment of the present application;

[0040] Figure 5 It is a flowchart of an object situation recognition method provided by an embodiment of the present application;

[0041] Figure 6 This is one of the structural schematic diagrams of a device for determining a situation recognition model provided in an embodiment of the present application;

[0042] Figure 7 This is a second structural diagram of a device for determining a situation recognition model provided in an embodiment of the present application;

[0043] Figure 8 It is a structural diagram of an object situation recognition model provided in an embodiment of the present application;

[0044] Fig. 9 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0046] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0047] The term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0048] In order to solve the problem of low accuracy of situation awareness in the prior art, using the method for determining the situation recognition model provided by the embodiment of the present application, the edge node can send the first model parameter to multiple base stations respectively, so that each base station can receive the first model parameter, and obtain multiple samples and sample labels of each sample, wherein each sample includes communication perception information of the first object, and the sample label of each sample is used to characterize the object situation of the first object. In this way, the base station can use the acquired multiple samples and the sample labels of each sample to train the situation recognition model, and the situation recognition model is determined based on the first model parameter to obtain a trained situation recognition model, and the trained situation recognition model includes a second model parameter, and sends a model parameter change to the edge node, and the model parameter change is the difference between the above-mentioned second model parameter and the above-mentioned first model parameter. In this way, the edge node can determine the updated first model parameter based on the multiple model parameter changes after receiving the model parameter changes sent by multiple base stations, and send the updated first model parameter to the multiple base stations respectively, so that each base station repeats the above process until the preset conditions are met to obtain an accurate situation recognition model.

[0049] Based on this, the object situation recognition method provided by the embodiment of the present application can obtain the communication perception information of the object to be recognized, and use the situation recognition model determined by the above process to process the communication perception information of the above model to be recognized to obtain the object situation of the object to be recognized. In this way, the accuracy of situation recognition can be effectively improved.

[0050] In order to facilitate a better understanding of the method for determining the situation recognition model provided in the embodiment of the present application in the future, before introducing the method for determining the situation recognition model provided in the embodiment of the present application, the architecture between the visual perception device, base station and edge node involved in the embodiment of the present application is first described in detail in conjunction with the accompanying drawings.

[0051] Figure 1 This is an architecture diagram between a visual perception device, a base station, and an edge node provided in an embodiment of the present application.

[0052] like Figure 1 As shown, the architecture diagram includes an edge node 10, multiple base stations 20, and multiple visual perception devices 30. Among them, the edge node 10 is connected to multiple base stations 20, each of the above base stations 20 is connected to multiple visual perception devices 30, and the multiple visual perception devices 30 are all visual perception devices deployed within the coverage range of the base station 20. It should be noted here that the above edge node 10 can be connected to multiple base stations through a transmission device, and the transmission device is not limited too much here.

[0053] It should also be noted that the edge node 10 is used to integrate the model parameters used to construct the situation recognition model. The base station 20 is used to construct the situation recognition model based on the model parameters provided by the edge node and perform training. The visual perception device 30 refers to related equipment such as a high-definition camera for collecting visual perception information, and the visual perception information is used to label samples.

[0054] based on Figure 1The architecture between the visual perception device, the base station and the edge node shown in the figure, the edge node 10 can send the first model parameters to the multiple base stations 20 connected to the edge node 10, so that each base station 20 can obtain multiple samples and the sample label of each sample after receiving the first model parameters sent by the edge node, and the sample label of each sample is annotated based on the visual perception information obtained by the multiple visual perception devices 30 connected to each base station 20, and the situation recognition model is trained using the multiple samples and the sample label of each sample to obtain the trained situation model. Since the above-mentioned situation recognition model can be constructed based on the first model parameters, the above The trained situation recognition model may include a second model parameter, so each base station 30 can determine the difference between the second model parameter and the first parameter as the model parameter change, and send the model parameter change to the above-mentioned edge node 10. In this way, the edge node 10 can determine the updated first model parameter based on the model parameter change received from multiple base stations 30, and send the updated first model parameter to the multiple base stations 30 respectively, so that each base station 30 can return to execute the above-mentioned steps of obtaining multiple samples and the sample label of each sample until the trained situation recognition model meets the preset jump to obtain an accurate target situation recognition model.

[0055] Through the above process, a more accurate target situation recognition model can be obtained. Each of the above base stations can obtain the communication perception information of the object to be identified, and use the target situation recognition model obtained through the above process to process the communication perception information of the object to be identified to obtain the object situation of the object to be identified.

[0056] based on Figure 1 The architecture between the visual perception device, base station and edge node shown in the figure is combined with the attached Figure 2 and Figure 3 The method for determining the situation recognition model applied to the base station provided in the embodiment of the present application is described in detail.

[0057] Figure 2 It is a flow chart of a method for determining a situation recognition model provided in an embodiment of the present application.

[0058] like Figure 2 As shown, the execution subject of the method may be a base station. Based on this, the method may specifically include the following steps:

[0059] S210: Receive a first model parameter sent by an edge node.

[0060] Among them, the first model parameter can be a relevant parameter used to construct a situation recognition model. The first model parameter can be one or more, and is not specifically limited here.

[0061] S220, obtaining multiple samples and a sample label of each sample.

[0062] In some embodiments, each of the above samples may include communication perception information of the first object, and the sample label of each sample is used to characterize the object status of the first object.

[0063] The first object may be a pedestrian, a vehicle or other object, which is not specifically limited here, and the object state of the first object may refer to the object state of the first object. Exemplarily, the sample label may be used to characterize whether the object state is normal or abnormal, which is not specifically limited here.

[0064] It should be noted that the method for determining the situation recognition model provided in the embodiment of the present application does not specifically limit the execution order of the above S210 and S220.

[0065] S230, training a situation recognition model using multiple samples and a sample label of each sample to obtain a trained situation recognition model.

[0066] In some embodiments, the situation recognition model can be constructed based on the first model parameters. In addition, the trained situation recognition model can include the second model parameters.

[0067] It should be noted that the first model parameter and the second model parameter may be the same or different, and are not specifically limited here. It should also be noted that since the embodiment of the present application does not specifically limit the number of the first model parameters, similarly, the embodiment of the present application does not specifically limit the number of the second model parameters.

[0068] Specifically, after obtaining multiple samples and the sample label of each sample, the base station can train the situation recognition model constructed based on the above-mentioned first model parameters based on the multiple samples and the sample label of each sample, and then obtain the trained situation recognition model.

[0069] In addition, the above situation recognition model can include Convolutional Neural Network (CNN) and lightGBM (Light Gradient Boosting Machine), where CNN is used for feature extraction and lightGBM is used for target recognition + situation awareness classification. Specifically, after obtaining the sample, the CNN model can be used for feature extraction to reduce the dimension of the data and improve the overall computational efficiency of the model; then the sample is input into the LightGBM model for classification training.

[0070] It should also be noted that lightGBM is an efficient and scalable machine learning algorithm based on Gradient Boosted Decision Trees (GBDT). As a member of the GBDT framework and a successor to the XGB (eXtreme Gradient Boosting) algorithm, LGBM combines the advantages of the previous algorithms in the GBDT algorithm framework, including XGB, and makes a series of further optimizations on this basis. The core purpose of the LGBM algorithm is to solve the problem of low computational efficiency of the GBDT algorithm framework when processing massive data.

[0071] S240: Send the model parameter change to the edge node.

[0072] The model parameter variation is the difference between the second model parameter and the first model parameter.

[0073] Specifically, after obtaining the trained situation recognition model, since the trained situation recognition model includes the second model parameters, the base station can first determine the change between the two based on the second model parameters and the above-mentioned first model parameters, that is, the above-mentioned model parameter change amount, and then can send the model parameter change amount to the edge node.

[0074] S250, receiving the updated first model parameters sent by the edge node, and returning to execute the step of obtaining multiple samples and a sample label of each sample until the trained situation recognition model meets the preset conditions, thereby obtaining the target situation recognition model.

[0075] The updated first model parameters are determined by the edge node based on the model parameter changes sent by multiple base stations, and are not specifically limited here.

[0076] In addition, the above-mentioned preset conditions can be pre-set based on actual experience or circumstances. For example, the preset condition can refer to the number of updates of the situation recognition model (i.e., the number of updates of the first model parameter) reaching a preset number threshold, and the preset number threshold can be set based on actual experience or circumstances.

[0077] Specifically, after sending the model parameter change to the edge node, the base station can receive the updated first model parameter sent by the edge node, so that it can return to execute the above S220 until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model.

[0078] In one example, the number of iterations of the base station training situation recognition model can be preset to N, the learning rate η, and the number of updates of the situation recognition model R. Based on this, the edge node can send the first model parameter w to the base station.T , T=0,1,...,R-1. In this way, the base station can T Construct a situation recognition model and use the obtained multiple samples and sample labels for training. When the number of iterations reaches N, the trained situation recognition model is obtained. The model parameters of the trained situation recognition model are Where i is used to represent the i-th base station. Thus, the base station can be based on w i T 、w T Determine the amount of change in model parameters And sent to the edge node, so that the edge node determines ΔW for updating the first model parameter, and determines the updated first model parameter w t+1 is (w t +ΔW), and send it to the base station, and continue to repeat the above process until the update times of the situation recognition model reaches R, then the target situation recognition model is obtained.

[0079] In an embodiment of the present application, the base station is capable of receiving a first model parameter sent by an edge node, and obtaining multiple samples and a sample label for each sample, each sample including communication perception information of a first object, and the sample label for each sample is used to characterize the object situation of the first object, and then the situation recognition model constructed based on the above-mentioned first model parameter can be trained using multiple samples and each sample label to obtain a trained situation recognition model, which includes a second model parameter, and then sending a model parameter change to the edge node, which is the difference between the second model parameter and the first model parameter, that is, characterizing the change before and after the training of the above-mentioned situation recognition model, and receiving the updated first model parameter sent by the edge node, and repeating the above process until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model. Since the above-mentioned updated first model parameters are determined by the edge node based on the model parameter changes sent by multiple base stations, that is, after the base station completes training of the situation recognition model each time, the edge node can determine the model parameters used to reconstruct the situation recognition model based on the changes before and after the training of the situation recognition model of each base station, so that each base station continues to train the situation recognition model until the trained situation recognition model meets the preset conditions. In this way, a more accurate target situation recognition model can be obtained, and then in the subsequent situation recognition process, the accuracy of situation recognition can be effectively improved.

[0080] In the method for determining the situation recognition model provided in the embodiment of the present application, it is necessary to train the situation recognition model constructed based on the first model parameter multiple times. Therefore, a large number of samples and sample labels of each sample need to be obtained in each training process for training the situation recognition model. In this way, in order to more accurately obtain samples and sample labels for training to obtain a more accurate situation recognition model, in one embodiment, Figure 3 As shown, the above S220 may specifically include the following steps:

[0081] S310: Acquire communication perception information of multiple first objects.

[0082] S320: Acquire visual perception information of multiple second objects.

[0083] The visual perception information of each second object may include the object posture of the second object.

[0084] S330: For each communication perception information of the first object, determine target visual perception information matching the communication perception information of the first object from visual perception information of multiple second objects.

[0085] S340, determining the communication perception information of the first object as a sample, and determining the object posture of the second object in the target visual perception information as a sample label of the sample.

[0086] Specifically, in the process of obtaining samples and sample labels, the base station can first obtain the communication perception information of multiple first objects, and obtain the visual perception information of multiple second objects. Since the visual perception information of each second object can include the object posture of the second object, the communication perception information of each first object can be determined from the visual perception information of multiple second objects. The target visual perception information that matches the communication perception information of the first object can then be determined, and the communication perception information of the first object can be determined as a sample, and the object posture of the second object in the target visual perception information can be determined as the sample label of the sample.

[0087] In this embodiment, by acquiring visual perception information including the object posture of the second object and determining target visual perception information that matches the communication perception information of each first object, sample labels can be annotated for the corresponding communication perception information based on the object posture of the second object included in the target visual perception information, thereby accurately obtaining multiple samples and the label of each sample, thereby facilitating the subsequent accurate training of the posture recognition model.

[0088] Because in the above embodiment, for each communication perception information of the first object, it is necessary to determine the target visual perception information that matches the communication perception information of the first object from multiple visual perception information, so that the sample label can be annotated for the above communication perception information based on the object posture of the second object contained in the target visual perception information. In this way, in order to accurately determine the target visual perception information corresponding to each communication perception information and facilitate the accurate representation of subsequent sample labels, in one embodiment, the communication perception information of each first object may include the first relative position of the first object relative to the base station, and the visual perception information of each second object may include the second object position of the second object relative to the base station. Based on this, the above 330 may specifically include the following steps:

[0089] For each visual perception information of the second object, determining a distance between the first relative position and the second relative position;

[0090] When the distance between the first relative position and the second relative position is less than a preset distance threshold, the visual perception information corresponding to the second relative position is determined to match the target visual perception information with the communication perception information of the first object.

[0091] The above-mentioned preset distance threshold may be preset based on actual experience or circumstances, and is not specifically limited here.

[0092] Specifically, for the visual perception information of each second object, since the communication perception information of the first object includes a first relative position, and the visual perception information of the second object includes a second relative position, the base station can first determine the distance between the first relative position and the second relative position, and then determine whether the distance is less than a preset distance threshold. If the distance is less than the preset distance threshold, it means that the first object and the second object are too close and are the same object. In this way, the visual perception information corresponding to the second relative position can be determined as the target visual perception information corresponding to the communication perception information.

[0093] It should be noted that if the distance between the above-mentioned first relative position and the second relative position is greater than or equal to the preset distance threshold, it indicates that the first object and the second object are too far apart and are different objects. Then, it can be confirmed that the visual perception information corresponding to the second relative position is not visual perception information that matches the communication perception information of the first object.

[0094] In this embodiment, based on the first relative position of the first object included in the communication perception information relative to the base station, and the second relative position of the second object included in the visual perception information relative to the base station, it is possible to determine whether the first object and the second object are the same object through the first relative position and the second relative position, and thus accurately determine the target visual perception information corresponding to each communication perception information, so that the communication perception information can be subsequently sampled based on the object posture of the second object included in the target visual perception information.

[0095] Based on this, in some embodiments, the above-mentioned first relative position may include at least one first position reference quantity in the horizontal arrival angle of the first object relative to the base station, the vertical arrival angle of the first object relative to the base station, and the signal transmission delay between the first object and the base station.

[0096] Likewise, the second relative position may include at least one second position reference quantity including a horizontal arrival angle of the second object relative to the base station, a vertical arrival angle of the second object relative to the base station, and a signal transmission delay between the second object and the base station.

[0097] In this embodiment, it is possible to accurately determine whether the corresponding first object and second object are the same object through at least one first position reference included in the first relative position and at least one second position reference included in the second relative position, thereby facilitating subsequent accurate sample labeling of the communication perception information based on the object posture of the second object included in the target visual perception information.

[0098] In the above embodiment, it is necessary to determine whether the first object and the second object are the same object based on the distance between the first relative position and the second relative position, and then perform sample annotation for the communication perception information. Based on this, in order to accurately determine whether the first object and the second object are the same object, so as to accurately and effectively perform sample annotation subsequently, in one embodiment, when the distance between the first relative position and the second relative position is less than a preset distance threshold, the visual perception information corresponding to the second relative position is determined. Before matching the target visual perception information with the communication perception information of the first object, the method for determining the situation recognition model provided in the embodiment of the present application may also include the following steps:

[0099] For each first position reference quantity in the first relative position, determine a position offset between the first position reference quantity and a second position reference quantity corresponding to the first position reference quantity in the second relative position, to obtain at least one position offset;

[0100] In a case where at least one position offset is smaller than a preset offset threshold, it is determined that the distance between the first relative position and the second relative position is smaller than a preset distance threshold.

[0101] Among them, the above-mentioned preset offset threshold can be set in advance based on actual experience or circumstances, and no specific limitation is made here. Moreover, the preset offset thresholds corresponding to different position offsets can be different, and no specific limitation is made here.

[0102] Specifically, since the first relative position can include at least one first position reference quantity, correspondingly, the second relative position can include at least one second position reference quantity. Therefore, the base station can determine the position offset between each first position reference quantity in the first relative position and the corresponding second position reference quantity in the second relative position. In this way, the base station can obtain at least one position offset, and then can determine whether all of the at least one position offset are less than the preset offset threshold. If so, it can be determined that the distance between the first relative position and the second relative position is less than the preset distance threshold, that is, it indicates that the first object and the second object are the same object.

[0103] It should be noted that if any one of the at least one position offset is greater than or equal to the above-mentioned preset offset threshold, it can be determined that the distance between the first relative position and the second relative position is greater than or equal to the preset distance threshold, that is, it indicates that the first object and the second object are different objects.

[0104] In one example, if the first relative position includes the horizontal arrival angle A1 of the first object relative to the base station, the vertical arrival angle B1 of the first object relative to the base station, and the signal transmission delay C1 between the first object and the base station. The second relative position includes the horizontal arrival angle A2 of the second object relative to the base station, the vertical arrival angle B2 of the second object relative to the base station, and the signal transmission delay C2 between the second object and the base station. Based on this, the base station can calculate the position offsets |A1 - A2|, |B1 - B2|, and |C1 - C2|. In this way, when |A1 - A2| < a, |B1 - B2| < b, and |C1 - C2| < c, it is determined that the distance between the first relative position and the second relative position is less than the preset distance threshold. Wherein, a, b, and c are the preset offset thresholds corresponding to each position offset respectively. In this embodiment, for each type of position reference quantity, the position offset between the first position reference quantity and the second position reference quantity can be calculated, and then it can be accurately determined whether the first object and the second object are the same object by determining whether at least one position offset is less than the corresponding preset offset threshold, so as to perform sample annotation accurately and effectively subsequently.

[0105] In another embodiment, the step of determining the distance between the first relative position and the second relative position may specifically include the following steps:

[0106] Based on the correspondence between the preset position reference quantity and the estimated position of the object, matching at least one first position reference quantity in the first relative position to obtain the estimated object position of the first object;

[0107] Based on the correspondence between the preset position reference quantity and the estimated position of the object, matching at least one second position reference quantity in the second relative position to obtain the estimated object position of the second object;

[0108] A distance between the estimated object position of the first object and the estimated object position of the second object is determined as the distance between the first relative position and the second relative position.

[0109] Among them, the correspondence between the above-mentioned preset position reference and the estimated position of the object can be pre-set based on actual experience or circumstances. For example, the correspondence between the preset position reference and the estimated position of the object can be the correspondence expressed by the objective function, which is a related function for calculating the estimated position of the object based on the position reference, which may depend on the specific situation and is not specifically limited here.

[0110] In addition, the estimated position of the object may refer to the geographical location of the object, which is not specifically limited herein.

[0111] Specifically, since the first relative position may include at least one first position reference, and the second relative position may include at least one second position reference, based on this, the base station can match at least one first position reference in the first relative position based on the correspondence between the preset position reference and the estimated position of the object, and obtain the estimated object position of the first object. Similarly, the base station can also match at least one second position reference in the second relative position based on the correspondence between the preset position reference and the estimated object position, and obtain the estimated object position of the second object. In this way, after the base station obtains the estimated object position of the first object and the estimated object position of the second object, it can directly determine the distance between the estimated object position of the first object and the estimated object position of the second object, which is the distance between the first relative position and the second relative position.

[0112] In this embodiment, the geographic location of the first object can be calculated based on at least one first position reference included in the first relative position, and the geographic location of the second object can be calculated based on at least one second position reference included in the second relative position, and based on the geographic locations of the first object and the second object, the distance between the first relative position and the second relative position can be accurately obtained, and then it can be accurately determined whether the first object and the second object are the same object, so as to facilitate subsequent accurate and effective sample labeling.

[0113] In addition, it should be noted that the communication perception information involved in the embodiments of the present application may include, in addition to the relative position of the above-mentioned object relative to the base station, the number of bits encoded for the object on the physical channel, the number of symbols modulated for the object on the physical channel, the difference in uplink and downlink time advance, the time-varying Doppler shift of the relative motion between the signal source and the object, the time advance between the signal source and the object, the radar scattering cross-sectional area, the radar scattering image, the power change value between the transmit power and the receive power, the cyclic prefix length, the allocated bandwidth part (Bandwidth Part, BWP) bandwidth, the starting frequency band position, the horizontal departure angle of the object relative to the base station, the vertical departure angle of the object relative to the base station, etc., which are not specifically limited here.

[0114] In order to describe the method for determining the situation recognition model provided in the embodiment of the present application in more detail, in one embodiment, the above S320 may specifically include the following steps:

[0115] Acquire multiple images and image information of each image through a visual perception device;

[0116] Obtaining relative position information of the visual perception device relative to the base station;

[0117] For each of the plurality of images, based on the object position of the second object in the image and the object reduction ratio, calculate relative position information between the second object and the visual perception device;

[0118] Determine a second object position of the second object relative to the base station based on the relative position information of the visual perception device relative to the base station and the relative position information between the second object and the visual perception device;

[0119] Among them, the above-mentioned visual perception device can be one or more, and the embodiment of the present application does not specifically limit the number of the visual perception devices.

[0120] In addition, each of the above images may include a second object, and accordingly, the image information of each image may include the object position of the second object in the image, the object reduction ratio, and the object posture. The above object reduction ratio is the ratio between the actual size of the object and the size of the object in the image, and the above object posture may be obtained by recognizing the above image using a preset image recognition model, which is not specifically limited here.

[0121] Specifically, the base station can obtain multiple images and image information of each image through the visual perception device, and obtain the relative position information of the visual perception device relative to the base station. Since each of the above images contains a second object, the image information of each image may include the object position of the second object in the image, the object reduction ratio and the object posture. Based on this, the base station can calculate the relative position information between the second object and the visual perception device for each image based on the object position of the second object in the image and the object reduction ratio. In this way, after obtaining the relative position information of the visual perception device relative to the base station and the relative position information between the second object and the visual perception device, the base station can determine the second object position of the second object relative to the base station based on the above two relative position information.

[0122] It should be noted that in the above-mentioned process of obtaining visual perception information through the visual perception device, the visual perception device can collect visual perception information in real time or periodically, and send the collected visual perception information to the corresponding base station through the fifth generation mobile communication technology (5th Generation Mobile Communication Technology, 5G) network or customer terminal equipment (Customer Premise(s) Equipment, CPE).

[0123] It should also be noted that, since the base station needs to obtain visual perception information through visual perception equipment, it is necessary to deploy visual perception equipment in the area where object situation recognition is required before this step. Therefore, the visual perception equipment can be deployed according to the following rules: First, the coverage area of ​​the base station may include at least one cell, and the visual perception equipment can be deployed at the antenna position of each cell. In this way, a small number of visual perception equipment can be deployed compared to the number of deployments required by the existing technology, thereby reducing costs. Second, the visual perception equipment can be deployed in scenes such as residential areas, commercial areas, squares, parks, and hospitals. Third, the visual perception equipment is required to be deployed in a position where objects such as vehicles and pedestrians can be photographed. Fourth, the visual perception equipment is deployed at a similar position to the antenna panel of the corresponding cell, and the inclination angle is the same as the inclination angle of the antenna panel to obtain a similar perception range.

[0124] In this embodiment, the relative position information of the second object relative to the visual perception device can be calculated through multiple images acquired by the visual perception device and the image information of each image. Combined with the acquired relative position relationship of the visual perception device relative to the base station, the second relative position of the second object relative to the base station can be accurately determined.

[0125] The following is combined with Figure 4The method for determining the situation recognition model applied to the edge node provided in the embodiment of the present application is described in detail.

[0126] Figure 4 It is a flow chart of a method for determining a situation recognition model provided in an embodiment of the present application.

[0127] like Figure 4 As shown, the execution subject of the method may be an edge node. Based on this, the method for determining the situation recognition model provided in the embodiment of the present application may specifically include the following steps:

[0128] S410, sending the first model parameters to the multiple base stations respectively, so that each base station obtains multiple samples and a sample label of each sample, and trains the situation recognition model using the multiple samples and the sample label of each sample to obtain a trained situation recognition model.

[0129] Each sample may include communication perception information of the first object, and correspondingly, the sample label of each sample may be used to characterize the object state of the first object.

[0130] In addition, the situation recognition model can be constructed based on the first model parameters. The trained situation recognition model can include the second model parameters.

[0131] Specifically, since the edge node is connected to multiple base stations, the edge node can send the first model parameters to the multiple base stations respectively, so that each base station can obtain multiple samples and the sample label of each sample, and use the obtained multiple samples and the sample label of each sample to train the situation recognition model constructed based on the above-mentioned first model parameters to obtain a trained situation recognition model.

[0132] S420: Receive model parameter changes respectively sent by multiple base stations.

[0133] The model parameter variation is the difference between the second model parameter and the first model parameter.

[0134] Specifically, after each base station has trained a situation recognition model including the second model parameter, each base station can determine the difference between the second model parameter and the first model parameter, that is, the model parameter change amount, and then send the model parameter change amount to the edge node. In this way, the edge node can receive the model parameters sent by multiple base stations respectively.

[0135] S430, sending the updated first model parameters to multiple base stations respectively, so that each base station returns to execute the step of obtaining multiple samples and training labels for each sample, until the trained situation recognition model meets the preset conditions, and the target situation recognition model is obtained.

[0136] The updated first model parameter is determined by the edge node based on the model parameter variation sent by multiple base stations.

[0137] In addition, the above-mentioned preset condition can be preset based on actual experience or situation, and is not specifically limited here. For example, the preset condition can refer to the number of updates of the situation recognition model (i.e., the number of updates of the first model parameter) reaching a preset number threshold, and the preset number threshold can be set based on actual experience or situation.

[0138] Specifically, after obtaining the model parameter changes sent by multiple base stations respectively, the edge node determines the updated first model parameters based on the model parameter changes sent by the multiple base stations, and sends the updated first model parameters to the multiple base stations respectively, so that each base station can return to the step of obtaining multiple samples and the sample label of each sample after receiving the updated first model parameter, until the trained situation recognition model meets the preset conditions and obtains the target situation recognition model.

[0139] In an embodiment of the present application, the edge node can send the first model parameter to multiple base stations respectively, so that each base station can obtain multiple samples and the sample label of each sample, wherein each sample includes the communication perception information of the first object, and the sample label of each sample is used to characterize the object situation of the first object, and the situation recognition model constructed based on the first model parameter is trained using the multiple samples and the sample label of each sample to obtain the trained situation recognition model, and the trained situation recognition model includes the second model parameter. In this way, the edge node can receive the model parameter change amount sent by multiple base stations respectively, and the model parameter change amount is the difference between the second model parameter and the first model parameter, that is, it characterizes the change before and after the training of the above situation recognition model, and sends the updated first model parameter to multiple base stations respectively, so that each base station continues to repeatedly train the situation recognition model constructed based on the first model parameter until the trained situation recognition model meets the preset conditions, and the target situation recognition model is obtained. Since the above-mentioned updated first model parameters are determined by the edge node based on the model parameter changes sent by multiple base stations, that is, after the base station completes training of the situation recognition model each time, the edge node can determine the model parameters used to reconstruct the situation recognition model based on the changes before and after the training of the situation recognition model of each base station, so that each base station continues to train the situation recognition model until the trained situation recognition model meets the preset conditions. In this way, a more accurate target situation recognition model can be obtained, and then in the subsequent situation recognition process, the accuracy of situation recognition can be effectively improved.

[0140] In order to accurately describe the method for determining the situation recognition model provided in the embodiment of the present application, in one embodiment, before the above S430, the method for determining the situation recognition model provided in the embodiment of the present application may specifically include the following steps:

[0141] Calculate the average value of changes in multiple model parameters;

[0142] The first model parameters are updated based on the average value to obtain updated first model parameters.

[0143] Specifically, after receiving the model parameter changes respectively sent by multiple base stations, the edge node can calculate the average value of the multiple model parameter changes, and then update the first model parameter based on the average value to obtain the updated first model parameter. More specifically, the edge node can determine that the sum of the above average value and the first model parameter is the updated first model parameter.

[0144] In another embodiment, before the above S430, the method for determining the situation recognition model provided in the embodiment of the present application may further include the following steps:

[0145] Perform weighted summation on multiple model parameter changes based on preset weights to obtain model parameter update;

[0146] The first model parameter is updated based on the model parameter update amount to obtain the updated first model parameter.

[0147] The above-mentioned preset weights may be preset based on actual experience or circumstances, and are not specifically limited here.

[0148] Specifically, after receiving the model parameter change amounts respectively sent by multiple base stations, the edge node can perform weighted summation of the multiple model parameter change amounts based on preset weights to obtain a model parameter update amount, and then can update the first model parameter based on the model parameter update amount to obtain an updated first model parameter. More specifically, the edge node can determine the sum of the above model parameter update amount and the above first model parameter as the updated first model parameter.

[0149] In this embodiment, after receiving the model parameter changes sent by multiple base stations respectively, the edge node can comprehensively understand the situation of training the situation recognition model of each base station, and update the original first model parameters by calculating the average of multiple model parameter changes or taking the weighted sum of multiple model parameter changes to obtain more accurate first model parameters, so as to facilitate the subsequent training of a more accurate situation recognition model.

[0150] Based on this, in one embodiment, the preset weights may include sub-weights corresponding to a plurality of model parameter changes, respectively, and the sub-weights corresponding to each model parameter change may be determined based on the ratio between the number of samples corresponding to the model parameter change and the total number of samples, wherein the total number of samples may refer to the sum of the numbers corresponding to the plurality of model parameter changes. It should also be noted that the number of samples corresponding to the model parameter change refers to the number of samples used to train a situation recognition model that can obtain the model parameter change.

[0151] In this embodiment, the sub-weight corresponding to each model parameter change can be accurately determined based on the number of samples of the situation recognition model trained by each base station, so that the model parameter changes can be accurately weighted and summed to obtain a more accurate first model parameter.

[0152] Based on the method for determining the situation recognition model provided in the embodiment of the present application, not only can a more accurate model for situation recognition be obtained to improve the accuracy of subsequent situation recognition, but also a variety of software and hardware resources of the smart community can be integrated to propose the following: Figure 1 The overall architecture shown reduces costs and transmission delays. In addition, in order to address the problem that existing methods are extremely dependent on high-definition cameras in community scenes to collect data on people or objects, which has high construction costs and requires a large amount of transmission bandwidth resources, the existing image-based situation awareness method can be used to annotate the samples obtained by the embodiment of the present application, and train the situation recognition model. It is also possible to fully integrate and utilize the training data resources of various scenes in the community and the distributed computing resources of the base station. In addition, the situation recognition model provided in the embodiment of the present application can realize comprehensive mining and analysis of feature data, input lightGBM for target recognition and situation awareness, reduce latency and improve reasoning efficiency.

[0153] The following is combined with Figure 5 The object situation recognition method provided in the embodiment of the present application is described in detail.

[0154] Figure 5 It is a flow chart of an object situation recognition method provided in an embodiment of the present application.

[0155] like Figure 5 As shown, the execution subject of the method may be a base station. Based on this, the method may specifically include the following steps:

[0156] S510: Acquire communication perception information of the object to be identified.

[0157] S520, using Figures 2 to 4The target situation recognition model determined by the steps involved in the flowchart of any situation recognition model determination method shown processes the communication perception information of the object to be recognized to obtain the object situation of the object to be recognized.

[0158] The above-mentioned objects to be identified may refer to pedestrians, vehicles and other objects, which are not specifically limited here.

[0159] Specifically, the base station can obtain the communication perception information of the object to be identified, and then can use the method steps shown in any of the above embodiments to determine the target situation recognition model, process the communication perception information of the object to be identified, and obtain the object situation of the object to be identified.

[0160] In an embodiment of the present application, the communication perception information of the object to be identified can be obtained, and the communication perception information of the object to be identified can be processed by obtaining a target situation recognition model using the above-mentioned situation recognition model determination method to obtain the object situation of the object to be identified. After the base station has trained the situation recognition model each time, the edge node can determine the model parameters for reconstructing the situation recognition model based on the changes before and after the situation recognition model of each base station is trained, so that each base station continues to train the situation recognition model until the trained situation recognition model meets the preset conditions, and then the above-mentioned target situation recognition model can be obtained. In this way, the accuracy of situation recognition can be effectively improved in the process of object situation recognition through the target situation recognition model.

[0161] Based on the same inventive concept, the embodiment of the present application provides a method for determining a situation recognition model, which is applied to a base station, specifically in combination with the attached Figure 6 The device for determining the situation recognition model provided in the embodiment of the present application is described in detail.

[0162] Figure 6 It is a structural diagram of a device for determining a situation recognition model provided in an embodiment of the present application.

[0163] like Figure 6 As shown, the situation recognition model determination device 600 may include:

[0164] A receiving module 610, configured to receive a first model parameter sent by an edge node;

[0165] An acquisition module 620 is used to acquire multiple samples and a sample label of each sample, each sample including communication perception information of the first object; the sample label of each sample is used to characterize the object state of the first object;

[0166] A training module 630 is used to train a situation recognition model using multiple samples and a sample label of each sample, wherein the situation recognition model is constructed based on the first model parameter to obtain a trained situation recognition model, wherein the trained situation recognition model includes the second model parameter;

[0167] A sending module 640 is used to send a model parameter change amount to an edge node, where the model parameter change amount is a difference between the second model parameter and the first model parameter;

[0168] The receiving module 610 is used to receive the updated first model parameters sent by the edge node, where the updated first model parameters are determined by the edge node based on the model parameter changes sent by multiple base stations, and return to execute the step of obtaining multiple samples and the sample label of each sample until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model.

[0169] In one embodiment, the apparatus for determining the situation recognition model provided in the embodiment of the present application may further include:

[0170] An acquisition module, specifically used to acquire communication perception information of multiple first objects;

[0171] An acquisition module, specifically used to acquire visual perception information of a plurality of second objects, wherein the visual perception information of each second object includes an object state of the second object;

[0172] a determination module, configured to determine, for each communication perception information of the first object, target visual perception information matching the communication perception information of the first object from visual perception information of a plurality of second objects;

[0173] The determination module is further used to determine the communication perception information of the first object as a sample, and to determine the object posture of the second object in the target visual perception information as a sample label of the sample.

[0174] In one embodiment, the communication perception information of each first object includes a first relative position of the first object relative to the base station; the visual perception information of each second object includes a second relative position of the second object relative to the base station; based on this, the above-mentioned determination module is specifically used to:

[0175] For each visual perception information of the second object, determining a distance between the first relative position and the second relative position;

[0176] When the distance between the first relative position and the second relative position is less than a preset distance threshold, the visual perception information corresponding to the second relative position is determined to match the target visual perception information with the communication perception information of the first object.

[0177] In one embodiment, the first relative position includes at least one first position reference quantity of a horizontal arrival angle of the first object relative to the base station, a vertical arrival angle of the first object relative to the base station, and a signal transmission delay between the first object and the base station;

[0178] The second relative position includes a horizontal arrival angle of the second object relative to the base station, a vertical arrival angle of the second object relative to the base station, and at least one second position reference quantity of a signal transmission delay between the second object and the base station.

[0179] In one embodiment, the determination module is further configured to determine, for each first position reference in the first relative position, a position offset between the first position reference and a second position reference corresponding to the first position reference in the second relative position, to obtain at least one position offset;

[0180] The above-mentioned determination module is also used to determine that the distance between the first relative position and the second relative position is less than a preset distance threshold when at least one position offset is less than a preset offset threshold.

[0181] In one embodiment, the apparatus for determining the situation recognition model provided in the embodiment of the present application further includes:

[0182] A matching module, configured to match at least one first position reference quantity in the first relative position based on a correspondence between a preset position reference quantity and an estimated object position, so as to obtain an estimated object position of the first object;

[0183] The matching module is further used to match at least one second position reference quantity in the second relative position based on the corresponding relationship between the preset position reference quantity and the estimated position of the object to obtain the estimated object position of the second object;

[0184] The accurate module is specifically used to determine the distance between the estimated object position of the first object and the estimated object position of the second object, which is the distance between the first relative position and the second relative position.

[0185] In one embodiment, the apparatus for determining the situation recognition model provided in the embodiment of the present application may further include:

[0186] An acquisition module, specifically used to acquire multiple images and image information of each image through a visual perception device, each image includes a second object, and the image information of each image includes an object position of the second object in the image, an object reduction ratio, and an object posture;

[0187] An acquisition module, specifically used to obtain relative position information of the visual perception device relative to the base station;

[0188] a calculation module, configured to calculate, for each of the plurality of images, relative position information between the second object and the visual perception device based on the object position of the second object in the image and the object reduction ratio;

[0189] The determination module is also used to determine the second object position of the second object relative to the base station based on the relative position information of the visual perception device relative to the base station and the relative position information between the second object and the visual perception device.

[0190] In an embodiment of the present application, the base station is capable of receiving a first model parameter sent by an edge node, and obtaining multiple samples and a sample label for each sample, each sample including communication perception information of a first object, and the sample label for each sample is used to characterize the object situation of the first object, and then the situation recognition model constructed based on the above-mentioned first model parameter can be trained using multiple samples and each sample label to obtain a trained situation recognition model, which includes a second model parameter, and then sending a model parameter change to the edge node, which is the difference between the second model parameter and the first model parameter, that is, characterizing the change before and after the training of the above-mentioned situation recognition model, and receiving the updated first model parameter sent by the edge node, and repeating the above process until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model. Since the above-mentioned updated first model parameters are determined by the edge node based on the model parameter changes sent by multiple base stations, that is, after the base station completes training of the situation recognition model each time, the edge node can determine the model parameters used to reconstruct the situation recognition model based on the changes before and after the training of the situation recognition model of each base station, so that each base station continues to train the situation recognition model until the trained situation recognition model meets the preset conditions. In this way, a more accurate target situation recognition model can be obtained, and then in the subsequent situation recognition process, the accuracy of situation recognition can be effectively improved.

[0191] Each module in the device for determining the situation recognition model provided in the embodiment of the present application can be implemented Figure 2 or Figure 3 The method steps of the illustrated embodiment can achieve the corresponding technical effects, and for the sake of brevity, they will not be repeated here.

[0192] Based on the same inventive concept, the embodiment of the present application also provides a device for determining a situation recognition model, which can be applied to edge nodes, specifically in combination with Figure 7 The device for determining the situation recognition model provided in the embodiment of the present application is described in detail.

[0193] Figure 7 It is a structural diagram of a device for determining a situation recognition model provided in an embodiment of the present application.

[0194] like Figure 7 As shown, the situation recognition model determination device 700 may include:

[0195] A sending module 710 is used to send the first model parameters to the multiple base stations respectively, so that each base station obtains multiple samples and a sample label of each sample, each sample includes communication perception information of the first object; the sample label of each sample is used to characterize the object situation of the first object, and the situation recognition model is trained using the multiple samples and the sample label of each sample, the situation recognition model is constructed based on the first model parameters, and a trained situation recognition model is obtained, and the trained situation recognition model includes a second model parameter;

[0196] The receiving module 720 is used to receive the model parameter change amounts respectively sent by multiple base stations, where the model parameter change amount is the difference between the second model parameter and the first model parameter;

[0197] The sending module 710 is also used to send the updated first model parameters to multiple base stations respectively. The updated first model parameters are determined by the edge node based on the model parameter changes sent by the multiple base stations, so that each base station returns to execute the step of obtaining multiple samples and the sample label of each sample until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model.

[0198] In one embodiment, the apparatus for determining the situation awareness model provided in the embodiment of the present application may further include:

[0199] A calculation module, used for calculating the average value of changes in multiple model parameters;

[0200] An updating module, used for updating the first model parameter based on the average value to obtain an updated first model parameter;

[0201] or,

[0202] A weighted summation module is used to perform weighted summation on multiple model parameter changes based on preset weights to obtain a model parameter update amount;

[0203] The updating module is further used to update the first model parameter based on the model parameter update amount, and the updated first model parameter.

[0204] In an embodiment of the present application, the edge node can send the first model parameter to multiple base stations respectively, so that each base station can obtain multiple samples and the sample label of each sample, wherein each sample includes the communication perception information of the first object, and the sample label of each sample is used to characterize the object situation of the first object, and the situation recognition model constructed based on the first model parameter is trained using the multiple samples and the sample label of each sample to obtain the trained situation recognition model, and the trained situation recognition model includes the second model parameter. In this way, the edge node can receive the model parameter change amount sent by multiple base stations respectively, and the model parameter change amount is the difference between the second model parameter and the first model parameter, that is, it characterizes the change before and after the training of the above situation recognition model, and sends the updated first model parameter to multiple base stations respectively, so that each base station continues to repeatedly train the situation recognition model constructed based on the first model parameter until the trained situation recognition model meets the preset conditions, and the target situation recognition model is obtained. Since the above-mentioned updated first model parameters are determined by the edge node based on the model parameter changes sent by multiple base stations, that is, after the base station completes training of the situation recognition model each time, the edge node can determine the model parameters used to reconstruct the situation recognition model based on the changes before and after the training of the situation recognition model of each base station, so that each base station continues to train the situation recognition model until the trained situation recognition model meets the preset conditions. In this way, a more accurate target situation recognition model can be obtained, and then in the subsequent situation recognition process, the accuracy of situation recognition can be effectively improved.

[0205] Each module in the device for determining the situation recognition model provided in the embodiment of the present application can be implemented Figure 4 The method steps of the illustrated embodiment can achieve the corresponding technical effects, and for the sake of brevity, they will not be repeated here.

[0206] Based on the same inventive concept, the embodiment of the present application further provides an object situation recognition device, which can be applied to a base station, specifically in combination with Figure 8 The object situation recognition device provided in the embodiment of the present application is described in detail.

[0207] Figure 8 It is a structural diagram of an object situation recognition device provided in an embodiment of the present application.

[0208] like Figure 8 As shown, the object situation recognition model 800 may include:

[0209] An acquisition module 810 is used to acquire communication perception information of an object to be identified;

[0210] Processing module 820, for using Figures 2 to 4The target situation recognition model determined by the steps involved in the flowchart of any situation recognition model determination method shown processes the communication perception information of the object to be recognized to obtain the object situation of the object to be recognized.

[0211] In an embodiment of the present application, the communication perception information of the object to be identified can be obtained, and the communication perception information of the object to be identified can be processed by obtaining a target situation recognition model using the above-mentioned situation recognition model determination method to obtain the object situation of the object to be identified. After the base station has trained the situation recognition model each time, the edge node can determine the model parameters for reconstructing the situation recognition model based on the changes before and after the situation recognition model of each base station is trained, so that each base station continues to train the situation recognition model until the trained situation recognition model meets the preset conditions, and then the above-mentioned target situation recognition model can be obtained. In this way, the accuracy of situation recognition can be effectively improved in the process of object situation recognition through the target situation recognition model.

[0212] Each module in the object situation recognition device provided in the embodiment of the present application can be implemented Figure 5 The method steps of the illustrated embodiment can achieve the corresponding technical effects, and for the sake of brevity, they will not be repeated here.

[0213] Fig. 9 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application is shown.

[0214] The electronic device may include a processor 901 and a memory 902 storing computer program instructions.

[0215] Specifically, the processor 901 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0216] The memory 902 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 902 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 902 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 902 may be inside or outside the electronic device. In a particular embodiment, the memory 902 is a non-volatile solid-state memory.

[0217] The memory may include read-only memory (ROM), random access memory (RAM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical or other physical / tangible memory storage devices. Thus, typically, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to an aspect of the present disclosure.

[0218] The processor 901 reads and executes computer program instructions stored in the memory 902 to implement any one of the situation recognition model determination methods or object situation recognition methods in the above embodiments.

[0219] In one example, the electronic device may further include a communication interface 903 and a bus 910. Fig. 9 As shown, the processor 901, the memory 902, and the communication interface 903 are connected via a bus 910 and communicate with each other.

[0220] The communication interface 903 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0221] Bus 910 includes hardware, software or both, and the parts of electronic equipment are coupled to each other.For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 910 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.

[0222] In addition, in combination with the method for determining the situation recognition model or the method for object situation recognition in the above embodiments, the embodiments of the present application may provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by the processor, the method for determining the situation recognition model or the method for object situation recognition provided in the embodiments of the present application is implemented.

[0223] An embodiment of the present application also provides a computer program product, which includes a computer program, and the computer program is executed by a processor to implement the method for determining a situation recognition model or the method for object situation recognition as provided in the embodiment of the present application.

[0224] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0225] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0226] The above are only specific implementation methods of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A method for determining a situation recognition model, characterized in that: Applied to a base station, the method comprises: Receiving a first model parameter sent by an edge node; Acquire a plurality of samples and a sample label of each of the samples, each of the samples comprising communication perception information of a first object; the sample label of each of the samples is used to characterize the object state of the first object; Using the multiple samples and the sample labels of each of the samples to train a situation recognition model, the situation recognition model is constructed based on the first model parameters to obtain a trained situation recognition model, the trained situation recognition model includes a second model parameter; Sending a model parameter change amount to the edge node, where the model parameter change amount is a difference between the second model parameter and the first model parameter; Receive the updated first model parameter sent by the edge node, where the updated first model parameter is determined by the edge node based on the model parameter change sent by multiple base stations, and return to execute the step of obtaining multiple samples and a sample label for each of the samples until the trained situation recognition model meets the preset conditions, thereby obtaining the target situation recognition model.

2. The method according to claim 1, characterized in that The obtaining of a plurality of samples and a sample label of each of the samples comprises: Acquire communication perception information of multiple first objects; Acquire visual perception information of a plurality of second objects, wherein the visual perception information of each second object includes an object state of the second object; For each communication perception information of the first object, determining target visual perception information matching the communication perception information of the first object from the visual perception information of the plurality of second objects; The communication perception information of the first object is determined as a sample, and the object posture of the second object in the target visual perception information is determined as a sample label of the sample.

3. The method according to claim 2, characterized in that The communication perception information of each first object includes a first relative position of the first object relative to the base station; the visual perception information of each second object includes a second relative position of the second object relative to the base station; The determining, from the visual perception information of the plurality of second objects, target visual perception information matching the communication perception information of the first object comprises: For each visual perception information of the second object, determining a distance between the first relative position and the second relative position; In the case where the distance between the first relative position and the second relative position is less than a preset distance threshold, visual perception information corresponding to the second relative position is determined to match target visual perception information with communication perception information of the first object.

4. The method according to claim 3, characterized in that The first relative position includes at least one first position reference quantity of a horizontal arrival angle of the first object relative to the base station, a vertical arrival angle of the first object relative to the base station, and a signal transmission delay between the first object and the base station; The second relative position includes a horizontal arrival angle of the second object relative to the base station, a vertical arrival angle of the second object relative to the base station, and at least one second position reference quantity of a signal transmission delay between the second object and the base station.

5. The method according to claim 4, characterized in that Before determining the visual perception information corresponding to the second relative position when the distance between the first relative position and the second relative position is less than a preset distance threshold, and matching the target visual perception information with the communication perception information of the first object, the method further includes: For each of the first position reference quantities in the first relative position, determine a position offset between the first position reference quantity and a second position reference quantity corresponding to the first position reference quantity in the second relative position, to obtain at least one position offset; In a case where the at least one position offset is smaller than a preset offset threshold, it is determined that the distance between the first relative position and the second relative position is smaller than a preset distance threshold.

6. The method according to claim 4, characterized in that The determining the distance between the first relative position and the second relative position comprises: Based on the correspondence between the preset position reference quantity and the estimated position of the object, matching at least one first position reference quantity in the first relative position to obtain the estimated position of the object of the first object; Based on the correspondence between the preset position reference quantity and the estimated object position, matching at least one second position reference quantity in the second relative position to obtain the estimated object position of the second object; A distance between the estimated object position of the first object and the estimated object position of the second object is determined as the distance between the first relative position and the second relative position.

7. The method according to claim 3, characterized in that The obtaining of visual perception information of a plurality of second objects includes: Acquire multiple images and image information of each image through a visual perception device, each image includes a second object, and the image information of each image includes an object position, an object reduction ratio, and an object posture of the second object in the image; Acquiring relative position information of the visual perception device relative to the base station; For each of the plurality of images, based on an object position of a second object in the image and a reduction ratio of the object, calculating relative position information between the second object and the visual perception device; Based on the relative position information of the visual perception device relative to the base station and the relative position information between the second object and the visual perception device, the second object position of the second object relative to the base station is determined.

8. A method for determining a situation recognition model, characterized in that: Applied to an edge node, the method comprises: Sending first model parameters to multiple base stations respectively, so that each base station obtains multiple samples and a sample label of each sample, each sample includes communication perception information of a first object; the sample label of each sample is used to characterize the object situation of the first object, and using the multiple samples and the sample label of each sample to train a situation recognition model, the situation recognition model is constructed based on the first model parameters, and a trained situation recognition model is obtained, and the trained situation recognition model includes a second model parameter; receiving model parameter change amounts respectively sent by the multiple base stations, the model parameter change amount being a difference between the second model parameter and the first model parameter; The updated first model parameters are respectively sent to the multiple base stations, where the updated first model parameters are determined by the edge node based on the model parameter changes sent by the multiple base stations, so that each base station returns to execute the step of obtaining multiple samples and a sample label for each of the samples until the trained situation recognition model meets the preset conditions, thereby obtaining the target situation recognition model.

9. The method according to claim 8, characterized in that Before sending the updated first model parameters to the multiple base stations respectively, the method further includes: Calculate the average value of changes in multiple model parameters; updating the first model parameter based on the average value to obtain an updated first model parameter; or, Perform weighted summation on multiple model parameter changes based on preset weights to obtain model parameter update; The first model parameter is updated based on the model parameter update amount to obtain the updated first model parameter.

10. A method for object situation recognition, characterized in that: Applied to a base station, the method comprises: Acquire communication perception information of the object to be identified; The communication perception information of the object to be identified is processed using the target situation recognition model as described in any one of claims 1 to 9 to obtain the object situation of the object to be identified.

11. A device for determining a situation recognition model, characterized in that: Applied to a base station, the device comprises: A receiving module, used for receiving a first model parameter sent by an edge node; An acquisition module, configured to acquire a plurality of samples and a sample label of each of the samples, each of the samples comprising communication perception information of a first object; and the sample label of each of the samples is used to characterize the object state of the first object; A training module, used to train a situation recognition model using the multiple samples and the sample label of each of the samples, wherein the situation recognition model is constructed based on the first model parameters to obtain a trained situation recognition model, wherein the trained situation recognition model includes a second model parameter; A sending module, used for sending a model parameter change amount to the edge node, where the model parameter change amount is a difference between the second model parameter and the first model parameter; The receiving module is also used to receive the updated first model parameter sent by the edge node, where the updated first model parameter is determined by the edge node based on the model parameter change sent by multiple base stations, and return to execute the step of obtaining multiple samples and the sample label of each of the samples until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model.

12. A device for determining a situation recognition model, characterized in that: Applied to an edge node, the device comprises: A sending module, configured to send first model parameters to a plurality of base stations respectively, so that each base station acquires a plurality of samples and a sample label of each of the samples, wherein each of the samples includes communication perception information of a first object; the sample label of each of the samples is used to characterize the object situation of the first object, and a situation recognition model is trained using the plurality of samples and the sample label of each of the samples, wherein the situation recognition model is constructed based on the first model parameters to obtain a trained situation recognition model, wherein the trained situation recognition model includes a second model parameter; A receiving module, configured to receive model parameter changes respectively sent by the multiple base stations, wherein the model parameter change is a difference between the second model parameter and the first model parameter; The sending module is also used to send updated first model parameters to the multiple base stations respectively, where the updated first model parameters are determined by the edge node based on the model parameter changes sent by the multiple base stations, so that each base station returns to execute the step of obtaining multiple samples and a sample label for each of the samples until the trained situation recognition model meets the preset conditions to obtain the target situation recognition model.

13. An object situation recognition device, characterized in that: Applied to a base station, the base station comprises: An acquisition module, used to acquire communication perception information of an object to be identified; A processing module is used to process the communication perception information of the object to be identified using the target situation recognition model as described in any one of claims 1 to 9 to obtain the object situation of the object to be identified.

14. An electronic device, characterized in that: The device comprises: a processor and a memory storing computer program instructions; The processor reads and executes the computer program instructions to implement the method for determining the situation recognition model according to any one of claims 1 to 9 or the method for object situation recognition according to claim 10.

15. A computer storage medium, characterized in that The computer storage medium stores computer program instructions, which, when executed by a processor, implement the method for determining a situation recognition model as described in any one of claims 1 to 9 or the method for object situation recognition as described in claim 10.

16. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the method for determining a situation recognition model as described in any one of claims 1 to 9 or the method for object situation recognition as described in claim 10.

Images