Data processing method and apparatus for optimizing target detection models

Abstract

This application discloses a data processing method and apparatus for optimizing an object detection model. The method includes: acquiring scene data to be processed, which represents data representing environmental perception interactions in an object detection scene; performing incremental learning processing on the scene data to be processed based on first scene features to obtain updated scene feature data, where the first scene features are scene features from a previous object detection scene; performing model update processing on the updated scene feature data based on a first object detection model to obtain an updated object detection model, where the first object detection model is an object detection model used in the previous object detection scene; and performing model performance evaluation processing on the updated object detection model to obtain a second object detection model. By progressively learning and updating data from different scenes, the method achieves the learning and updating of the object detection model in new scenes, thereby improving the real-time performance and accuracy of object detection in new scenes and environments.

Description

Technical Field

[0001] This application relates to the field of computers, and more specifically, to a data processing method and apparatus for optimizing target detection models. Background Technology

[0002] In recent years, object detection models have been widely used in various fields to achieve object detection in various scenarios. However, existing object detection models mainly collect labeled sample data and train the model based on the sample data. For target data with different data distributions than the sample data, the object detection model has difficulty in effectively detecting and recognizing it. For new data, the model needs to be retrained from scratch, which is time-consuming and inefficient.

[0003] Military scenarios present unique challenges, such as variable environments, diverse targets, and high real-time requirements. Military target detection tasks typically require processing large amounts of real-time data. When drones or robots perform tasks in military scenarios, they need to achieve adaptive target detection and subsequent decision-making based on different scenario data, which is difficult to achieve with existing model detection technologies.

[0004] Therefore, this application proposes a data processing method and apparatus for continuous optimization of target detection models. Summary of the Invention

[0005] The main objective of this application is to provide a data processing method and apparatus for optimizing target detection models, so as to solve one or more of the above-mentioned technical problems and achieve the technical effect of improving the real-time performance and accuracy of target detection.

[0006] To achieve the above objectives, the first aspect of this application proposes a data processing method for optimizing a target detection model, comprising: Acquire scene data to be processed, wherein the scene data to be processed is data used to represent the environmental perception interaction of the target detection scene; The scene data to be processed is subjected to incremental learning processing based on the first scene feature to obtain updated scene feature data, wherein the first scene feature is the scene feature of the previous target detection scene; The updated scene feature data is subjected to model update processing based on the first target detection model to obtain an updated target detection model, wherein the first target detection model is a target detection model used in the previous target detection scene; The updated target detection model is subjected to model performance evaluation to obtain a second target detection model, wherein the second target detection model is the optimized target detection model.

[0007] Further, the updated scene feature data is subjected to model update processing based on the first target detection model to obtain the updated target detection model, including: Scene features are extracted from the scene data to be processed to obtain scene feature data to be processed; The scene feature data to be processed and the first scene feature data are subjected to scene similarity-based judgment processing. If the scene similarity between the scene feature data to be processed and the first scene feature data is greater than or equal to a preset similarity threshold, the first model update parameters are obtained. The first target detection model is then updated based on the updated scene feature data and the first model update parameters to obtain the updated target detection model. If the scene similarity between the scene feature data to be processed and the first scene feature data is less than a preset similarity threshold, a second model update parameter is obtained. The first target detection model is then updated based on the updated scene feature data and the second model update parameter to obtain the updated target detection model.

[0008] Further, the updated scene feature data is subjected to model update processing based on the first target detection model to obtain the updated target detection model, including: The first target detection model is optimized based on the model update parameters to obtain a scene-optimized first target detection model. The updated scene feature data is subjected to forward inference processing based on the scene to optimize the first target detection model, thereby obtaining the model update training set; Based on the model update training set and the scene optimization first target detection model, a joint loss function is constructed. The target detection model is then trained using the joint loss function to obtain the updated target detection model.

[0009] Further, a joint loss function is constructed based on the model-updated training set and the scene-optimized first object detection model. The object detection model is then trained using the joint loss function to obtain the updated object detection model, which includes: The scene-optimized first target detection model is subjected to associated feature extraction processing to obtain associated feature map data, wherein the associated feature map data is feature relationship map data used to represent the scene-optimized first target detection model; The training set of the model is updated, and the prediction response processing of the first target detection model based on the scene is optimized to obtain prediction response data; A joint loss function for the target detection model is constructed based on the associated feature map data and the predicted response data; The target detection model is subjected to model iterative training based on the joint loss function to obtain the updated target detection model, wherein the updated target detection model is used to represent the target detection model that satisfies the preset model convergence rule during iterative training.

[0010] Furthermore, acquiring the data for the scene to be processed includes: Acquire multimodal scene data, wherein the multimodal scene data is data used to represent the target detection scene environment collected by multiple sensors; Feature extraction processing is performed on the multimodal scene data to obtain multimodal scene feature data; The feature data of the multiple modal scenes are subjected to feature fusion-based data fusion processing to obtain the scene data to be processed.

[0011] Furthermore, after performing model performance evaluation on the updated target detection model to obtain the second target detection model, the method further includes: The second target detection model is used to perform target detection processing on the target detection scene for a preset time period to obtain target detection result data. The target detection result data is subjected to behavior feature extraction processing to obtain target behavior feature data, wherein the target behavior feature data is feature data used to represent the target behavior in the scene within a preset time period; The target behavior feature data scene anomaly detection processing is performed to obtain anomaly detection result data, wherein the anomaly detection result data is detection result data used to represent scene anomalies in the target detection scene; The anomaly detection results are subjected to online learning processing of the scene anomaly detection model to obtain an updated anomaly detection model.

[0012] Furthermore, after performing model performance evaluation on the updated target detection model to obtain the second target detection model, the method further includes: Obtain the task data to be updated, where the task data to be updated is the data used to represent the task to be updated in the object detection scene; The task data to be updated is processed by extracting the first task features to obtain the first task features to be updated, wherein the first task feature data to be updated is data used to represent static task features; The second task feature is extracted from the task data to be updated to obtain the second task feature to be updated, wherein the second task feature data is used to represent the dynamic task features. Incremental learning based on prior task features is performed on the first and second task features to be updated respectively to obtain updated task feature data; The updated task feature data is processed by model update based on the first task model to obtain the updated task model, where the task model can be a scene anomaly detection model. The updated task model is subjected to model performance evaluation to obtain the second task model, which is the updated task model.

[0013] According to a second aspect of this application, a data processing apparatus for optimizing a target detection model is proposed, comprising: The data acquisition module is used to acquire scene data to be processed, wherein the scene data to be processed is data representing the environmental perception interaction of the target detection scene; The incremental learning module is used to perform incremental learning processing on the scene data to be processed based on the first scene feature to obtain updated scene feature data, wherein the first scene feature is the scene feature of the previous target detection scene. The model update module is used to perform model update processing on the updated scene feature data based on the first target detection model to obtain an updated target detection model, wherein the first target detection model is a target detection model used in the previous target detection scene; The model evaluation module is used to perform model performance evaluation on the updated target detection model to obtain a second target detection model, wherein the second target detection model is the optimized target detection model.

[0014] According to a third aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing the computer to perform the above-described data processing method for optimizing a target detection model.

[0015] According to a fourth aspect of this application, an electronic device is proposed, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the at least one processor to perform the data processing method described above for target detection model optimization.

[0016] The technical solutions provided by the embodiments of this application may include the following beneficial effects: In this application, scene data to be processed is acquired, wherein the scene data to be processed is data used to represent the environmental perception interaction of the target detection scene; incremental learning processing based on a first scene feature is performed on the scene data to be processed to obtain updated scene feature data, wherein the first scene feature is the scene feature of the previous target detection scene; model update processing based on a first target detection model is performed on the updated scene feature data to obtain an updated target detection model, wherein the first target detection model is the target detection model used in the previous target detection scene; model performance evaluation processing is performed on the updated target detection model to obtain a second target detection model, wherein the second target detection model is the optimized target detection model. By progressively learning and updating data under different scenes, the target detection model for new scenes is learned and updated, thereby improving the real-time performance and accuracy of target detection in new scenes and environments. Attached Figure Description

[0017] The accompanying drawings, which form part of this application, are used to provide a further understanding of the application and to make other features, objects, and advantages of the application more apparent. The illustrative embodiments and descriptions of this application are used to explain the application and do not constitute an undue limitation of the application. In the drawings: Figure 1 A flowchart of a data processing method for optimizing a target detection model is provided in this application; Figure 2 A flowchart of a data processing method for optimizing a target detection model provided in this application; Figure 3 A flowchart of a data processing method for optimizing a target detection model is provided in this application; Figure 4 This is a schematic diagram of a data processing device for optimizing a target detection model, as provided in this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] In this application, the terms "upper," "lower," "left," "right," "front," "rear," "top," "bottom," "inner," "outer," "middle," "vertical," "horizontal," "lateral," and "longitudinal" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are primarily for the purpose of better describing this application and its embodiments, and are not intended to limit the indicated device, element, or component to having a specific orientation, or to be constructed and operated in a specific orientation.

[0021] Furthermore, in addition to indicating location or positional relationship, some of the aforementioned terms may also have other meanings. For example, the term "above" may also be used in some cases to indicate a certain dependency or connection relationship. Those skilled in the art can understand the specific meaning of these terms in this application based on the specific circumstances.

[0022] Furthermore, the terms "installation," "setup," "equipped with," "connection," "linked," and "socketing" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral structure; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or an internal connection between two devices, components, or parts. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0023] In an optional embodiment of this application, a data processing method for optimizing a target detection model is proposed. Figure 1 A flowchart of a data processing method for optimizing a target detection model is provided in this application, as shown below. Figure 1 As shown, the method includes the following steps: S101: Obtain the scene data to be processed; The scene data to be processed is data used to represent the environmental perception interaction in the target detection scene; In some optional embodiments of this application, a data processing method for optimizing a target detection model is proposed, including: Acquire multimodal scene data, which represents the target detection scene environment collected by multiple sensors; perform feature extraction processing on the multimodal scene data to obtain multiple modal scene feature data; perform data fusion processing based on feature fusion on the multiple modal scene feature data to obtain the scene data to be processed.

[0024] In optional embodiments of this application, multiple types of scene environment perception and interaction data are collected using various sensors. For example, scene image data is collected using a visual sensor, and scene infrared data and radar data are collected using an infrared sensor and a radar sensor, respectively. The multimodal scene data includes first-modal scene data, second-modal scene data, and third-modal scene data. Feature extraction processing is performed on the first-modal scene data to obtain first-modal scene feature data; feature extraction processing is performed on the second-modal scene data to obtain second-modal scene feature data; feature extraction processing is performed on the third-modal scene data to obtain third-modal scene feature data; and feature fusion processing based on scene features is performed on the first-modal scene data, second-modal scene data, and third-modal scene data to obtain the scene data to be processed. By utilizing multimodal data to construct a joint representation space, the technical effect of improving the accuracy of target detection in target detection scenarios is achieved.

[0025] S102: Perform incremental learning processing on the scene data to be processed based on the first scene features to obtain updated scene feature data; The first scene feature refers to the scene features in the previous object detection scene. The first scene feature is the scene feature in the first object detection model. The first object detection model can be an object detection model in the previous object detection scene, or it can be a pre-trained multimodal model CLIP. The pre-trained multimodal model can be a convolutional neural network pre-trained on ImageNet, such as ResNet or VGG. The incremental learning process based on the first scene feature for the scene data to be processed includes: extracting model features from the object detection model in the previous object detection scene to obtain the model features of the previous object detection model; and performing incremental learning based on the model features of the previous object detection model on the scene data to be processed to obtain updated scene feature data. Alternatively, if the first object detection model is a pre-trained multimodal model CLIP, the incremental learning process based on the first scene feature for the scene data to be processed includes: performing incremental learning based on the general features in the aforementioned pre-trained model on the scene data to be processed to obtain updated scene feature data.

[0026] S103: Perform model update processing on the updated scene feature data based on the first target detection model to obtain the updated target detection model; The first object detection model is an object detection model used in the previous object detection scenario; In an optional embodiment of this application, a data processing method for optimizing a target detection model is proposed, comprising: Scene features are extracted from the scene data to be processed to obtain scene feature data. The scene feature data to be processed includes the labeled dataset of the scene to be detected. The scene labeled dataset contains sample data of different scene features in the scene to be detected. The scene feature data includes scene environment feature data and scene target feature data. The scene environment feature data is used to represent the environment-related feature data in the target detection environment, such as lighting feature data, viewpoint feature data, occlusion feature data, etc. The scene target feature data is used to represent the feature data related to the target to be detected, such as feature data to represent the target type, target identifier, etc.

[0027] The scene feature data to be processed and the first scene feature data are processed based on scene similarity. If the scene similarity between the scene feature data to be processed and the first scene feature data is greater than or equal to a preset similarity threshold, the first model update parameters are obtained. The first target detection model is then updated based on the updated scene feature data and the first model update parameters to obtain the updated target detection model. If the scene similarity between the scene feature data to be processed and the first scene feature data is less than a preset similarity threshold, the second model update parameters are obtained. The first target detection model is then updated based on the updated scene feature data and the second model update parameters to obtain the updated target detection model.

[0028] In an optional embodiment of this application, by comparing the scene similarity between the scene to be processed and the scene in the prior scene, if the similarity between the scene to be processed and the prior scene is greater than or equal to a preset similarity threshold, the first model update parameters for the prior object detection model are determined, such as the number of fine-tuning layers for partial layer fine-tuning of the prior model, or fine-tuning the top or last few layers of the prior object detection model; if the similarity between the scene to be processed and the prior scene is less than the preset similarity threshold, the second model update parameters for the prior object detection model are determined, such as full-layer fine-tuning of the prior model, or fine-tuning all trainable layers of the prior object detection model.

[0029] In another optional embodiment of this application, the updated scene feature data is subjected to a judgment process based on a preset data volume threshold. If the updated scene feature data is less than the preset data volume threshold, the first model optimization parameters for the prior target detection model are determined, such as the number of fine-tuning layers for partial layer fine-tuning of the prior model, or fine-tuning the top or last few layers of the prior target detection model. If the updated scene feature data is greater than or equal to the preset data volume threshold, the second model optimization parameters for the prior target detection model are determined, such as full-layer fine-tuning of the prior model, or fine-tuning all trainable layers of the prior target detection model.

[0030] In another optional embodiment of this application, a data processing method for optimizing a target detection model is proposed. Figure 2 A flowchart of a data processing method for optimizing a target detection model is provided in this application, as shown below. Figure 2 As shown, the method includes the following steps: S201: Perform model optimization processing based on model update parameters on the first target detection model to obtain a scene-optimized first target detection model; S202: Perform forward inference processing on the updated scene feature data based on scene-optimized first target detection model to obtain the model update training set; S203: Based on the model update training set and scene optimization, construct a joint loss function for the first object detection model, and train the object detection model using the joint loss function to obtain the updated object detection model.

[0031] In another optional embodiment of this application, a data processing method for optimizing an object detection model is proposed, comprising: performing associated feature extraction processing on a scene-optimized first object detection model to obtain associated feature map data, wherein the associated feature map data is feature relationship graph data used to represent the scene-optimized first object detection model; performing prediction response processing on the model update training set based on the scene-optimized first object detection model to obtain prediction response data; constructing a joint loss function of the object detection model according to the associated feature map data and the prediction response data; and performing model iterative training processing on the object detection model based on the joint loss function to obtain an updated object detection model, wherein the updated object detection model is an object detection model used to represent the iterative training satisfying a preset model convergence rule.

[0032] In an optional embodiment of this application, the updated target detection model is trained by adopting the above method. The updated target detection model is learned by learning the relationship structure between model features from the prior target detection model and training the updated target detection model based on the updated scene feature data.

[0033] S104: Perform model performance evaluation on the updated target detection model to obtain the second target detection model.

[0034] The second object detection model is the optimized object detection model.

[0035] In another optional embodiment of this application, a data processing method for optimizing a target detection model is proposed. After obtaining a second target detection model, Figure 3 A flowchart of a data processing method for optimizing a target detection model is provided in this application, as shown below. Figure 3 As shown, the method includes the following steps: S301: Perform target detection processing on the target detection scene according to the second target detection model for a preset time period to obtain target detection result data; S302: Perform behavioral feature extraction processing on the target detection result data to obtain target behavioral feature data; Target behavior feature data refers to feature data used to represent the target behavior in a scene within a preset time period; S303: Process the scene anomaly detection data of the target behavior feature data to obtain anomaly detection result data; Anomaly detection result data refers to the detection result data used to represent scene anomalies in the target detection scene; S304: Perform online learning processing on the anomaly detection results to obtain an updated anomaly detection model.

[0036] In another optional embodiment of this application, a data processing method for optimizing a target detection model is proposed, comprising: Obtain the task data to be updated, where the task data to be updated is the data used to represent the task to be updated in the object detection scene; The task data to be updated is processed by extracting the first task features to obtain the first task features to be updated, wherein the first task feature data to be updated is data used to represent static task features; The second task feature is extracted from the task data to be updated to obtain the second task feature to be updated, wherein the second task feature data is used to represent the dynamic task features. Incremental learning based on prior task features is performed on the first and second task features to be updated respectively to obtain updated task feature data; The updated task feature data is processed by model update based on the first task model to obtain the updated task model, where the task model can be a scene anomaly detection model. The updated task model is subjected to model performance evaluation to obtain the second task model, which is the updated task model.

[0037] For example, when the task model is a scene anomaly detection model, the task data to be updated is data used to represent the abnormal behavior and abnormal target to be updated. The abnormal behavior and abnormal target to be updated are incrementally learned based on the prior anomaly detection model to obtain the updated task feature data; the updated task feature data is processed by model update based on the first scene anomaly detection model to obtain the updated anomaly detection model; the updated anomaly detection model is processed by model performance evaluation to obtain the second scene anomaly detection model.

[0038] In the embodiments of this application, by continuously learning the characteristics of new abnormal behaviors and abnormal targets, the accuracy of anomaly detection is improved, and real-time detection of abnormal behaviors and abnormal targets in the target detection scenario is achieved.

[0039] In another optional embodiment of this application, a data processing apparatus for optimizing a target detection model is proposed. Figure 4 A schematic diagram of a data processing device for optimizing a target detection model provided in this application is shown below. Figure 4 As shown, the device includes: The data acquisition module 41 is used to acquire scene data to be processed, wherein the scene data to be processed is data used to represent the environmental perception interaction of the target detection scene; The incremental learning module 42 is used to perform incremental learning processing on the scene data to be processed based on the first scene feature to obtain updated scene feature data, wherein the first scene feature is the scene feature of the previous target detection scene. The model update module 43 is used to perform model update processing on the updated scene feature data based on the first target detection model to obtain an updated target detection model, wherein the first target detection model is a target detection model used in the previous target detection scene; The model evaluation module 44 is used to perform model performance evaluation on the updated target detection model to obtain a second target detection model, wherein the second target detection model is the optimized target detection model.

[0040] The specific methods of execution of each unit in the above embodiments have been described in detail in the embodiments of the method, and will not be elaborated here.

[0041] In summary, this application involves: acquiring scene data to be processed, wherein the scene data to be processed is data representing the environmental perception interaction of the target detection scene; performing incremental learning processing on the scene data to be processed based on a first scene feature to obtain updated scene feature data, wherein the first scene feature is the scene feature of the previous target detection scene; performing model update processing on the updated scene feature data based on a first target detection model to obtain an updated target detection model, wherein the first target detection model is the target detection model used in the previous target detection scene; and performing model performance evaluation processing on the updated target detection model to obtain a second target detection model, wherein the second target detection model is the optimized target detection model. By progressively learning and updating data under different scenarios, the target detection model for new scenarios is learned and updated, thereby improving the real-time performance and accuracy of target detection in new scenarios and environments.

[0042] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0043] Obviously, those skilled in the art should understand that the various units or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device, or fabricating them separately as individual integrated circuit modules, or fabricating multiple modules or steps into a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0044] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A data processing method for optimizing a target detection model, characterized in that, include: Acquire scene data to be processed, wherein the scene data to be processed is data used to represent the target detection scene environment perception interaction, and the scene data to be processed includes scene image data, scene infrared data and scene radar data; The scene data to be processed is subjected to incremental learning processing based on the first scene feature to obtain updated scene feature data, wherein the first scene feature is the scene feature of the previous target detection scene; The updated scene feature data is subjected to model update processing based on the first target detection model to obtain an updated target detection model, wherein the first target detection model is a target detection model used in the previous target detection scene; The updated scene feature data is subjected to model update processing based on the first target detection model to obtain the updated target detection model, which includes: Scene features are extracted from the scene data to be processed to obtain scene feature data to be processed; The scene feature data to be processed and the first scene feature data are subjected to scene similarity-based judgment processing. If the scene similarity between the scene feature data to be processed and the first scene feature data is greater than or equal to a preset similarity threshold, a first model update parameter is obtained. The first model update parameter is used to fine-tune the top or last few layers of the prior target detection model. The first target detection model is updated according to the updated scene feature data and the first model update parameter to obtain the updated target detection model. If the scene similarity between the scene feature data to be processed and the first scene feature data is less than a preset similarity threshold, a second model update parameter is obtained. The second model update parameter is used to fine-tune all trainable layers of the prior target detection model. The first target detection model is updated based on the updated scene feature data and the second model update parameter to obtain the updated target detection model. The updated target detection model is subjected to model performance evaluation to obtain a second target detection model, wherein the second target detection model is the optimized target detection model.

2. The data processing method according to claim 1, characterized in that, The updated scene feature data is subjected to model update processing based on the first target detection model to obtain the updated target detection model, which includes: The first target detection model is optimized based on the model update parameters to obtain a scene-optimized first target detection model. The updated scene feature data is subjected to forward inference processing based on the scene to optimize the first target detection model, thereby obtaining the model update training set; Based on the model update training set and the scene optimization first target detection model, a joint loss function is constructed. The target detection model is then trained using the joint loss function to obtain the updated target detection model.

3. The data processing method according to claim 2, characterized in that, Based on the model update training set and the scene optimization first object detection model, a joint loss function is constructed. The object detection model is then trained using the joint loss function to obtain the updated object detection model, which includes: The scene-optimized first target detection model is subjected to associated feature extraction processing to obtain associated feature map data, wherein the associated feature map data is feature relationship map data used to represent the scene-optimized first target detection model; The training set of the model is updated, and the prediction response processing of the first target detection model based on the scene is optimized to obtain prediction response data; A joint loss function for the target detection model is constructed based on the associated feature map data and the predicted response data; The target detection model is subjected to model iterative training based on the joint loss function to obtain the updated target detection model, wherein the updated target detection model is used to represent the target detection model that satisfies the preset model convergence rule during iterative training.

4. The data processing method according to claim 1, characterized in that, The data to be processed in the scenario includes: Acquire multimodal scene data, wherein the multimodal scene data is data used to represent the target detection scene environment collected by multiple sensors; Feature extraction processing is performed on the multimodal scene data to obtain multimodal scene feature data; The feature data of the multiple modal scenes are subjected to feature fusion-based data fusion processing to obtain the scene data to be processed.

5. The data processing method according to claim 1, characterized in that, After performing model performance evaluation on the updated target detection model to obtain the second target detection model, the method further includes: The second target detection model is used to perform target detection processing on the target detection scene for a preset time period to obtain target detection result data. The target detection result data is subjected to behavior feature extraction processing to obtain target behavior feature data, wherein the target behavior feature data is feature data used to represent the target behavior in the scene within a preset time period; The target behavior feature data scene anomaly detection processing is performed to obtain anomaly detection result data, wherein the anomaly detection result data is detection result data used to represent scene anomalies in the target detection scene; The anomaly detection results are subjected to online learning processing of the scene anomaly detection model to obtain an updated anomaly detection model.

6. The data processing method according to claim 1, characterized in that, After performing model performance evaluation on the updated target detection model to obtain the second target detection model, the method further includes: Obtain the task data to be updated, where the task data to be updated is the data used to represent the task to be updated in the object detection scene; The task data to be updated is processed by extracting the first task features to obtain the first task features to be updated, wherein the first task feature data to be updated is data used to represent static task features; The second task feature is extracted from the task data to be updated to obtain the second task feature to be updated, wherein the second task feature data is used to represent the dynamic task features. Incremental learning based on prior task features is performed on the first and second task features to be updated respectively to obtain updated task feature data; The updated task feature data is processed by model update based on the first task model to obtain the updated task model, where the task model can be a scene anomaly detection model. The updated task model is subjected to model performance evaluation to obtain the second task model, which is the updated task model.

7. A data processing apparatus for optimizing a target detection model, characterized in that, include: The data acquisition module is used to acquire scene data to be processed, wherein the scene data to be processed is data used to represent the environmental perception interaction of the target detection scene, and the scene data to be processed includes scene image data, scene infrared data and scene radar data; The incremental learning module is used to perform incremental learning processing on the scene data to be processed based on the first scene feature to obtain updated scene feature data, wherein the first scene feature is the scene feature of the previous target detection scene. The model update module is used to perform model update processing on the updated scene feature data based on the first target detection model to obtain an updated target detection model, wherein the first target detection model is a target detection model used in the previous target detection scene; The updated scene feature data is subjected to model update processing based on the first target detection model to obtain the updated target detection model, which includes: Scene features are extracted from the scene data to be processed to obtain scene feature data to be processed; The scene feature data to be processed and the first scene feature data are subjected to scene similarity-based judgment processing. If the scene similarity between the scene feature data to be processed and the first scene feature data is greater than or equal to a preset similarity threshold, a first model update parameter is obtained. The first model update parameter is used to fine-tune the top or last few layers of the prior target detection model. The first target detection model is updated according to the updated scene feature data and the first model update parameter to obtain the updated target detection model. If the scene similarity between the scene feature data to be processed and the first scene feature data is less than a preset similarity threshold, a second model update parameter is obtained. The second model update parameter is used to fine-tune all trainable layers of the prior target detection model. The first target detection model is updated based on the updated scene feature data and the second model update parameter to obtain the updated target detection model. The model evaluation module is used to perform model performance evaluation on the updated target detection model to obtain a second target detection model, wherein the second target detection model is the optimized target detection model.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the data processing method for optimizing a target detection model as described in any one of claims 1-6.

9. An electronic device, characterized in that, include: At least one processor; The at least one processor is also connected in communication with a memory, wherein the memory stores a computer program that can be executed by the at least one processor to cause the at least one processor to perform the data processing method for target detection model optimization as described in any one of claims 1-6.

Images