Target identification method, electronic equipment and computer readable storage medium

Abstract

The invention discloses a target recognition method, electronic equipment and a computer readable storage medium, and the method comprises the steps: obtaining a to-be-recognized image and a corresponding recognition indication text, inputting the to-be-recognized image and the recognition indication text into a detection model, and obtaining a target recognition result outputted by the detection model; wherein the identification indication text comprises a to-be-identified category of a target in the target scene; classifying the target recognition results based on standard samples in the registered base library to obtain correct samples and misinformation samples; wherein the standard sample is retrieved from a registration base library based on the identification indication text, and the category corresponding to each target is correspondingly matched with one registration base library; and obtaining a corrected misinformation sample, performing fine tuning on the detection model by using a training sample composed of the corrected misinformation sample and the correct sample to obtain a fine-tuned detection model, and adding the training sample to the registration base library of the target corresponding category. According to the scheme, the accuracy of target identification can be improved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a target recognition method, an electronic device and a computer-readable storage medium. Background Art

[0002] As one of the most important tasks in the field of computer vision, the development of cross-modal target recognition technology has laid the foundation for downstream computer tasks. The cross-modal target recognition task is mainly to combine multiple different modal information to perform target recognition tasks and give the corresponding category of the image target. The current cross-modal target recognition scheme is limited by factors such as high system complexity, high dependence on training data quality, and long model training time, resulting in a high false detection rate of target categories. In view of this, how to improve the accuracy of target recognition has become an urgent problem to be solved. Summary of the invention

[0003] The main technical problem solved by the present application is to provide a target recognition method, an electronic device and a computer-readable storage medium, which can improve the accuracy of target recognition.

[0004] To solve the above technical problems, the first aspect of the present application provides a target recognition method, including: obtaining an image to be recognized and its corresponding recognition instruction text, inputting the image to be recognized and the recognition instruction text into a detection model, and obtaining a target recognition result output by the detection model; wherein the recognition instruction text includes the category to be recognized of the target in the target scene; classifying the target recognition result based on the standard samples in the registration base library to obtain correct samples and false alarm samples; wherein the standard samples are retrieved from the registration base library based on the recognition instruction text, and each category corresponding to the target is matched with a corresponding registration base library; obtaining the corrected false alarm samples, fine-tuning the detection model using the training samples composed of the corrected false alarm samples and the correct samples, obtaining the fine-tuned detection model, and adding the training samples to the registration base library of the category corresponding to the target.

[0005] To solve the above technical problem, the second aspect of the present application provides an electronic device, comprising a memory and a processor coupled to each other, wherein the memory stores program instructions, and the processor is used to execute the program instructions to implement the method described in the first aspect.

[0006] In order to solve the above technical problem, the third aspect of the present application provides a computer-readable storage medium, which stores program instructions that can be executed by a processor, and the program instructions are used to implement the method described in the first aspect above.

[0007] The above scheme obtains an image to be identified and its corresponding identification instruction text, wherein the identification instruction text includes the category to be identified of the target in the target scene, inputs the image to be identified and the identification instruction text into the detection model, obtains the target recognition result output by the detection model, and each target corresponding to a category is matched with a registration base library, retrieves a standard sample from the registration base library based on the identification instruction text, classifies the target recognition result based on the standard sample, obtains a correct sample and a false positive sample, corrects the false positive sample, obtains the corrected false positive sample, forms a training sample with the corrected false positive sample and the correct sample, fine-tunes the detection model using the training sample to obtain a fine-tuned detection model, and adds the training sample to the registration base library of the target corresponding category, fine-tunes the detection model using the training sample composed of the corrected false positive sample and the correct sample, can timely update the detection model, and add the training sample to the registration base library as the standard sample of the target corresponding category, forming a closed-loop optimization training system with a small number of samples, so that the recognition accuracy of the fine-tuned detection model is improved, thereby improving the accuracy of target recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them: Figure 1 It is a flowchart of an implementation method of the target recognition method of the present application; Figure 2 yes Figure 1 Part of the content in step S101 corresponds to a flowchart of an implementation method; Figure 3 It is a structural schematic diagram of an embodiment of the electronic device of the present application; Figure 4 It is a structural schematic diagram of an implementation method of a computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0009] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments, and different implementation methods can be adaptively combined. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0010] The terms "system" and "network" are often used interchangeably in this article. The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship. In addition, "many" in this article means two or more than two.

[0011] See also Figure 1 , Figure 1 : is a flow chart of an implementation method of the target recognition method of the present application. The target recognition method includes: S101: Obtain an image to be identified and its corresponding identification instruction text, input the image to be identified and the identification instruction text into a detection model, and obtain a target recognition result output by the detection model; wherein the identification instruction text includes the category to be identified of the target in the target scene.

[0012] Specifically, an image to be identified and its corresponding identification instruction text are obtained, wherein the identification instruction text includes the category to be identified of the target in the target scene, and the image to be identified and the identification instruction text are input into the detection model to obtain the target recognition result output by the detection model.

[0013] In one application, an image to be identified and recognition instruction text input together with the image to be identified are obtained, the recognition instruction text includes the category to be identified of the target in the target scene, the image to be identified and the recognition instruction text are input into a detection model, and a target recognition result output by the detection model is obtained.

[0014] In another application method, an image to be identified is obtained and a reference image corresponding to the image to be identified is determined, and matching recognition instruction text is generated based on the reference image. The recognition instruction text includes the category to be identified of the target in the target scene, and the recognition instruction text corresponds to the image to be identified. The image to be identified and the corresponding recognition instruction text are input into a detection model to obtain a target recognition result output by the detection model.

[0015] In an application scenario, an image to be identified and corresponding recognition instruction text manually input by the user are obtained. The recognition instruction text includes the category to be identified of the target in the target scene, which is used to indicate the target that the user wants to identify. The image to be identified and the recognition instruction text are input into a detection model to obtain a target recognition result output by the detection model.

[0016] In another application scenario, an image to be identified is obtained and a reference image corresponding to the image to be identified is determined. The reference image is processed using a deep neural network model to generate recognition instruction text that matches the reference image. The recognition instruction text includes the category to be identified of the target in the target scene, and the recognition instruction text corresponds to the image to be identified. The image to be identified and the corresponding recognition instruction text are input into a detection model to obtain a target recognition result output by the detection model.

[0017] In a specific application scenario, the target scene may be a scene containing power equipment, and the identification instruction text may be related to the power equipment. For example, when there is a plastic bag on a utility pole, the identification instruction text may be “Please identify the plastic bag on the utility pole”, or when cracks appear on the equipment, the identification instruction text may be “Please identify the defect in the shape of a crack on the equipment”, etc.

[0018] In some implementation scenarios, each image to be identified includes multiple categories, and the identification indication text is used to indicate the categories to be identified corresponding to multiple different targets that the user wants to identify.

[0019] In some implementation scenarios, each image to be identified includes multiple categories, and the identification indication text is used to identify the category to be identified corresponding to a single target that the subject wants to identify.

[0020] In some implementation scenarios, each image to be identified includes only a single category, and the identification indication text is used to indicate the category to be identified corresponding to the single target that the user wants to identify.

[0021] Optionally, a video stream may be acquired through a camera device disposed in the target scene, and a corresponding image to be identified may be acquired from the video stream, or an image to be identified actively input by a user may be acquired, and this application does not impose any specific limitation thereto.

[0022] S102: Classify the target recognition results based on the standard samples in the registration base database to obtain correct samples and false positive samples; wherein the standard samples are retrieved from the registration base database based on the recognition instruction text, and each target corresponding category is matched with a registration base database.

[0023] Specifically, each target category is matched with a registered base library, and a standard sample is retrieved from the registered base library based on the recognition instruction text. The target recognition result is classified based on the standard sample to obtain a correct sample and a false positive sample.

[0024] In one application, the standard sample and the target recognition result are input into an analysis model for scoring to obtain a comparison value output by the analysis model. When the output comparison value is greater than a preset threshold, the target recognition result is determined to be a correct sample. When the output comparison value is less than the preset threshold, the target recognition result is determined to be a false alarm sample.

[0025] In another application mode, the target recognition result is directly compared with the standard sample. If the target recognition result is completely consistent with the standard sample, the target recognition result is considered to be a correct sample, otherwise, the target recognition result is considered to be a false positive sample.

[0026] S103: Obtain the corrected false alarm samples, fine-tune the detection model using the training samples consisting of the corrected false alarm samples and the correct samples to obtain the fine-tuned detection model, and add the training samples to the registration base library of the target corresponding category.

[0027] Specifically, the false alarm samples are corrected to obtain corrected false alarm samples, the corrected false alarm samples and correct samples are combined into training samples, the detection model is fine-tuned using the training samples to obtain a fine-tuned detection model, and the training samples are added to the registered base library of the target category.

[0028] In one application method, the false alarm samples are corrected to obtain the corrected false alarm samples, the corrected false alarm samples and the correct samples are combined into training samples, all parameters of the detection model are fine-tuned using the training samples to obtain the fine-tuned detection model, and the training samples are added to the registered base library of the target category.

[0029] In another application method, an adapter module is provided between each level of the detection model to correct the false alarm samples, obtain the corrected false alarm samples, and form training samples with the corrected false alarm samples and correct samples. The adapters are fine-tuned using the training samples to obtain the fine-tuned detection model, and the training samples are added to the registered base library of the target category.

[0030] The above scheme obtains an image to be identified and its corresponding identification instruction text, wherein the identification instruction text includes the category to be identified of the target in the target scene, inputs the image to be identified and the identification instruction text into the detection model, obtains the target recognition result output by the detection model, and each target corresponding to a category is matched with a registration base library, retrieves a standard sample from the registration base library based on the identification instruction text, classifies the target recognition result based on the standard sample, obtains a correct sample and a false positive sample, corrects the false positive sample, obtains the corrected false positive sample, forms a training sample with the corrected false positive sample and the correct sample, fine-tunes the detection model using the training sample to obtain a fine-tuned detection model, and adds the training sample to the registration base library of the target corresponding category, fine-tunes the detection model using the training sample composed of the corrected false positive sample and the correct sample, can timely update the detection model, and add the training sample to the registration base library as the standard sample of the target corresponding category, forming a closed-loop optimization training system with a small number of samples, so that the recognition accuracy of the fine-tuned detection model is improved, thereby improving the accuracy of target recognition.

[0031] In one embodiment, the detection model includes a graphic-text alignment module. Before classifying the target recognition results based on the standard samples in the registered base library to obtain correct samples and false positive samples in step S101, it also includes: inputting the recognition indication text into the graphic-text alignment module, and using the graphic-text alignment module to retrieve the standard samples corresponding to the recognition indication text from the registered base library.

[0032] Specifically, the detection model includes an image-text alignment module, which is used to implement an image search function based on text description. The recognition indication text is input into the image-text alignment module and detected from the registered base library using the image-text alignment module to obtain a standard sample corresponding to the recognition indication text.

[0033] In one embodiment, see Figure 2 , Figure 2 yes Figure 1 Part of the content of step S101 corresponds to a flowchart of an implementation method. Among them, the detection model includes a matching module, a visual encoding module and a text encoding module. In step S101, an image to be identified and its corresponding identification instruction text are obtained, and the image to be identified and the identification instruction text are input into the detection model. The target recognition result output by the detection model specifically includes: S201: Acquire the image to be recognized and its corresponding recognition instruction text.

[0034] Specifically, an image to be identified and its corresponding identification indication text are obtained, wherein the identification indication text includes a category to be identified of a target in a target scene.

[0035] Optionally, the image to be recognized and the recognition indication text input together with the image to be recognized are obtained, or the image to be recognized is obtained and a reference image corresponding to the image to be recognized is determined, and matching recognition indication text is generated according to the reference image.

[0036] S202: Input the image to be recognized and the recognition indication text into the detection model, use the visual encoding module to obtain the visual features of the image to be recognized, and use the text encoding module to obtain the text features of the recognition indication text.

[0037] Specifically, the image to be recognized and the recognition indication text are input into the detection model, the visual encoding module is used to extract the visual features of the image to be recognized, and the text encoding module is used to extract the text features of the recognition indication text.

[0038] In one application scenario, the image to be recognized and the recognition instruction text are input into the detection model, and the visual encoding module is used to determine the target region of interest of the image to be recognized, and the feature vector of the image to be recognized is extracted from the target region of interest as the visual feature of the image to be recognized, and the target text label in the recognition instruction text is determined, and the feature vector of the recognition instruction text is extracted from the target text label as the text feature of the recognition instruction text.

[0039] It should be noted that in machine vision and image processing, the region of interest is the area to be processed in the image being processed, which is outlined in the form of a box, circle, ellipse, irregular polygon, etc. Various operators and functions are often used in machine vision software such as Halcon, OpenCV, and Matlab to obtain the region of interest and perform the next step of image processing.

[0040] S203: Obtain target recognition results based on visual features and text features using a matching module.

[0041] Specifically, the matching module of the detection model is used to obtain the target recognition result of the image to be detected based on the extracted visual features and text features. By fusing the two modal information of image and text, the target features in the target scene can be more comprehensively understood and characterized, effectively solving the ambiguity problem in single modality recognition and significantly improving the accuracy and precision of target recognition.

[0042] In an application scenario, a matching module is used to fuse the extracted visual features and text features. The fusion methods include but are not limited to splicing, addition, multiplication, etc. After the visual features and text features are fused together, a new comprehensive feature vector is formed. The vector contains information about both the image and the text. The similarity between the fused comprehensive feature vector and the known target category or sample is calculated. Among them, the similarity can be calculated using measurement methods such as cosine similarity and Euclidean distance. The degree of match between the predicted category in the image to be detected and each target category is determined according to the size of the similarity, thereby obtaining the target recognition result.

[0043] In this embodiment, in step S103, the detection model is fine-tuned using the training samples consisting of the corrected false alarm samples and the correct samples to obtain the fine-tuned detection model, which specifically includes: fine-tuning the matching module of the detection model using the training samples consisting of the corrected false alarm samples and the correct samples to obtain the fine-tuned detection model.

[0044] Specifically, the false positive samples are corrected and combined with the correct samples to form training samples, and the parameters of the matching module are fine-tuned to obtain a fine-tuned detection model.

[0045] It is understandable that the user can label and correct the false positive samples according to the true category of the target.

[0046] Optionally, the visual features of the image to be detected extracted by the visual encoding module can be directly used to fine-tune the parameters of the matching module, which can make the module fine-tuning faster and thus improve the accuracy of subsequent target recognition using the fine-tuned detection model.

[0047] Optionally, a time threshold condition can be set, and the detection model can be fine-tuned when the time threshold condition is met. The detection model can also be fine-tuned in time according to the recognition effect of the detection model, so as to obtain a timely updated detection model and accurate recognition results, greatly improving the service experience and service quality.

[0048] In another embodiment, the registered base database is configured with a priority that matches the category. In step S101, an image to be identified and its corresponding identification instruction text are obtained, and the image to be identified and the identification instruction text are input into the detection model to obtain a target recognition result output by the detection model. Specifically, the following steps are included: obtaining an image to be identified and its corresponding identification instruction text; inputting the image to be identified and the identification instruction text into the detection model to obtain a target recognition result output by the detection model according to the priority.

[0049] Specifically, an image to be recognized and recognition instruction text are obtained, wherein the recognition instruction text includes the category to be recognized of the target in the target scene, the image to be recognized and the recognition instruction text are input into the detection model, and the target recognition result output by the detection model according to the preset priority is obtained. By pre-setting the priority, the model can give priority to identifying and processing the most important targets, and outputting the results according to the priority can allow the detection model to concentrate more computing resources and time on processing high-priority targets, thereby improving the overall recognition efficiency of the model.

[0050] In a specific implementation scenario, the target scene is an electric power scene, and when there is oil leakage in the equipment and dangerous objects such as balloons or bird nests on the electric poles in the electric power scene, the priority of the equipment oil leakage is higher than the priority of the dangerous objects on the electric poles. Therefore, after the corresponding image to be identified and the identification instruction text are input into the detection model, the detection model will first output the recognition result of "equipment oil leakage" and then output the recognition result of "bird nest on the electric pole".

[0051] In this embodiment, the target recognition method further includes: generating a corresponding reporting event based on the target recognition result; wherein the reporting event corresponds to prompt information, and the prompt information is used to prompt that an abnormality occurs in the target in the target scene.

[0052] Specifically, a corresponding reporting event is generated according to the target recognition result. The reporting event corresponds to prompt information, and the prompt information is used to prompt the user that an abnormal situation has occurred in the target in the target scene, thereby ensuring timely response and processing of the emergency.

[0053] In one implementation scenario, the target identification result is pre-configured with a reporting frequency threshold corresponding to the priority, and the step of generating a corresponding reporting event based on the target identification result specifically includes: generating a reporting event in response to the reporting frequency of the target identification result satisfying the reporting frequency threshold.

[0054] Specifically, when the reporting frequency of the target recognition result meets a pre-configured reporting frequency threshold, a corresponding reporting event is generated.

[0055] In a specific implementation scenario, the target scenario is an electric power scenario, and when there is an oil leak in the equipment and dangerous objects such as balloons or bird nests on the electric poles in the electric power scenario, the reporting frequency threshold corresponding to the oil leak is that as long as the situation occurs once, a corresponding reporting event must be generated, thereby avoiding the occurrence of dangerous situations. For lower priority abnormal situations such as dangerous objects such as balloons or bird nests on electric poles, the corresponding reporting frequency threshold can be set to the maximum number of times within a fixed time period, for example, "a maximum of 5 data reports are allowed per hour", and only when this condition is met will the corresponding reporting event be generated, thereby avoiding waste of resources or excessive system burden caused by too frequent reporting.

[0056] In one embodiment, after step S103, the method further includes: in response to the category corresponding to the target recognition result being different from the category corresponding to the registered base library, generating a registered base library corresponding to a newly added category, and adding the target recognition result to the registered base library corresponding to the newly added category.

[0057] Specifically, when the category corresponding to the target recognition result is different from the category corresponding to the registered base database, a registered base database corresponding to the newly added category is generated, and the target recognition result is added to the registered base database corresponding to the newly added category. The new category can be trained by fine-tuning the model to improve the recognition accuracy of the new target category. At the same time, since the training samples used are the corresponding recognition results under the target scenario, a data business closed loop can be achieved to avoid information security issues caused by data leakage.

[0058] See also Figure 3 , Figure 3 : is a structural diagram of an embodiment of an electronic device of the present application, the electronic device 30 includes a memory 300 and a processor 302 coupled to each other, wherein the memory 300 stores program data (not shown), and the processor 302 calls the program data to implement the method in any of the above embodiments. For the description of the relevant content, please refer to the detailed description of the above method embodiment, which will not be repeated here. Specifically, the electronic device 30 includes but is not limited to: a desktop computer, a laptop computer, a tablet computer, a server, etc., which are not limited here. In addition, the processor 302 can also be called a CPU (Center Processing Unit). The processor 302 may be an integrated circuit chip with signal processing capabilities. The processor 302 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field-programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. In addition, the processor 302 can be implemented by an integrated circuit chip.

[0059] See also Figure 4 , Figure 4 It is a structural diagram of an implementation method of a computer-readable storage medium of the present application. The computer-readable storage medium 40 stores program data 400. When the program data 400 is executed by a processor, the method in any of the above embodiments is implemented. For descriptions of related contents, please refer to the detailed description of the above method embodiments, which will not be repeated here.

[0060] It should be noted that the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present implementation scheme.

[0061] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0062] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of each implementation method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., and other media that can store program codes.

[0063] The above are only implementation methods of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A target recognition method, characterized in that: include: Obtain an image to be identified and its corresponding identification instruction text, input the image to be identified and the identification instruction text into a detection model, and obtain a target recognition result output by the detection model; wherein the identification instruction text includes the category to be identified of the target in the target scene; The target recognition results are classified based on standard samples in the registration base database to obtain correct samples and false positive samples; wherein the standard samples are retrieved from the registration base database based on the recognition instruction text, and each category corresponding to the target is matched with a corresponding registration base database; The corrected false alarm samples are obtained, and the detection model is fine-tuned using training samples consisting of the corrected false alarm samples and the correct samples to obtain the fine-tuned detection model, and the training samples are added to the registered base library of the target corresponding category.

2. The method according to claim 1, characterized in that The detection model includes a picture-text alignment module, and before classifying the target recognition results based on the standard samples in the registered base library to obtain the correct samples and the false positive samples, it also includes: The identification instruction text is input into the image-text alignment module, and the image-text alignment module is used to retrieve the standard sample corresponding to the identification instruction text from the registered base database.

3. The method according to claim 1, characterized in that The detection model includes a matching module, a visual encoding module and a text encoding module. The step of obtaining an image to be recognized and its corresponding recognition indication text, inputting the image to be recognized and the recognition indication text into the detection model, and obtaining a target recognition result output by the detection model includes: Acquire the image to be recognized and its corresponding recognition indication text; Inputting the image to be recognized and the recognition indication text into the detection model, using the visual encoding module to obtain visual features of the image to be recognized, and using the text encoding module to obtain text features of the recognition indication text; The target recognition result is obtained by utilizing the matching module based on the visual features and the text features.

4. The method according to claim 3, characterized in that The method of fine-tuning the detection model using the training samples consisting of the corrected false positive samples and the correct samples to obtain the fine-tuned detection model includes: The matching module of the detection model is fine-tuned using the training samples consisting of the corrected false alarm samples and the correct samples to obtain the fine-tuned detection model.

5. The method according to claim 1, characterized in that The registered base database is configured with a priority that matches the category; The step of obtaining an image to be recognized and its corresponding recognition indication text, inputting the image to be recognized and the recognition indication text into a detection model, and obtaining a target recognition result output by the detection model includes: Acquire the image to be recognized and its corresponding recognition indication text; The image to be recognized and the recognition instruction text are input into the detection model to obtain the target recognition result output by the detection model according to the priority.

6. The method according to claim 5, characterized in that The method further comprises: A corresponding reporting event is generated based on the target recognition result; wherein the reporting event corresponds to prompt information, and the prompt information is used to prompt that an abnormality occurs in a target in the target scene.

7. The method according to claim 6, characterized in that The target identification result is pre-configured with a reporting frequency threshold corresponding to the priority; Generate a corresponding reporting event based on the target recognition result, including: In response to the reporting frequency of the target identification result satisfying the reporting frequency threshold, the reporting event is generated.

8. The method according to claim 1, characterized in that After obtaining the corrected false positive samples, fine-tuning the detection model using training samples consisting of the corrected false positive samples and the correct samples to obtain the fine-tuned detection model, and adding the training samples to the registration base library corresponding to the target category, the method further includes: In response to the category corresponding to the target recognition result being different from the category corresponding to the registered base library, a registered base library corresponding to a newly added category is generated, and the target recognition result is added to the registered base library corresponding to the newly added category.

9. An electronic device, characterized in that: The method comprises a memory and a processor coupled to each other, wherein the memory stores program instructions, and the processor is used to execute the program instructions to implement the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: Program instructions that can be executed by a processor are stored, and the program instructions are used to implement the method described in any one of claims 1 to 8.

Images