Multi-field image intelligent enhancement method based on memory and decision fusion

Abstract

The invention discloses a memory and decision fusion-based multi-field image intelligent enhancement method, device and equipment and a storage medium, and the method comprises the steps: responding to the input of first image data, and determining multiple pieces of target second image data in multiple pieces of second image data stored in a memory unit of a labeling reasoning model; determining adjustment data of each piece of target second image data through feature alignment operation from the first frequency domain of the first image data to the target second frequency domains of the multiple pieces of target second image data; and determining enhanced data of the first image data through fine adjustment of the determined adjustment data of the target second image data on the annotation reasoning model in the reasoning process. Through fusion of memory and decision, excellent performance in a multi-field complex scene is realized, the image enhancement precision and speed in a wide-view-field high-resolution scene are remarkably improved, and the problem of insufficient reasoning effect of an image annotation reasoning model in the multi-field complex scene is solved.

Description

Technical Field

[0001] The present application belongs to the field of image annotation, and specifically relates to a multi-domain image intelligent enhancement method, device, equipment and storage medium based on memory and decision fusion. Background Art

[0002] In the field of modern computer vision and image processing, image annotation and enhancement technologies are widely used in multiple industries and scenarios, including but not limited to medical image analysis, intelligent monitoring, unmanned driving, satellite remote sensing, etc. These application scenarios have high requirements for the accuracy and quality of image data, so efficient and accurate image annotation and enhancement technologies are needed to achieve reliable image processing effects in complex and changing environments.

[0003] At present, image annotation and enhancement technology mainly relies on a single distribution hypothesis, that is, assuming that the feature distribution of training data and inference data is consistent. At the same time, in order to improve the image enhancement effect, a method of model training based on high-quality annotated data is usually used.

[0004] However, due to the significant differences in resolution, texture, lighting, noise distribution, etc. between images in different fields, the model's generalization ability is insufficient in cross-domain tasks. At the same time, due to the high cost of obtaining high-quality labeled data, especially in multi-domain scenarios, the scarcity of data has significantly restricted the model training effect, which in turn affects the model's cross-domain generalization performance. Summary of the invention

[0005] The present application aims to provide a multi-domain image intelligent enhancement method, device, equipment and storage medium based on memory and decision fusion, at least to solve the problem of insufficient reasoning effect of image annotation reasoning model in complex scenarios in multiple fields.

[0006] In a first aspect, the present application discloses a multi-domain image intelligent enhancement method based on memory and decision fusion, including: In response to input of first image data, a plurality of target second image data are determined from a plurality of second image data stored in a memory unit of a labeled reasoning model; the reasoning model is used to determine enhanced data of the first image data for the input first image data according to the second image data stored in the memory unit; the first image data is image data in a first data domain to be entered, and the second image data is image data in a second data domain that has been entered; Determine adjustment data of each of the target second image data by performing a feature alignment operation from a first frequency domain of the first image data to a target second frequency domain of a plurality of the target second image data; The enhanced data of the first image data is determined by fine-tuning the labeled reasoning model during the reasoning process using the determined adjustment data of the target second image data.

[0007] In a second aspect, the embodiment of the present application further discloses a multi-domain image intelligent enhancement device based on memory and decision fusion, comprising: A reference selection module is used to determine a plurality of target second image data from a plurality of second image data stored in a memory unit of a labeled reasoning model in response to input of the first image data; the reasoning model is used to determine enhanced data of the first image data for the input first image data according to the second image data stored in the memory unit; the first image data is image data under a first data domain to be entered, and the second image data is image data under a second data domain that has been entered; an alignment adjustment module, configured to determine adjustment data of each of the target second image data by performing a feature alignment operation from a first frequency domain of the first image data to a target second frequency domain of a plurality of the target second image data; The reasoning enhancement module is used to determine the enhancement data of the first image data by fine-tuning the labeled reasoning model during the reasoning process through the determined adjustment data of the target second image data.

[0008] In a third aspect, an embodiment of the present application further discloses an electronic device, comprising a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the method described in the first aspect are implemented.

[0009] In a fourth aspect, an embodiment of the present application further discloses a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented.

[0010] In summary, in the embodiment of the present application, by determining multiple target second image data from multiple second image data stored in the memory unit of the annotated reasoning model, not only can the existing high-quality annotated data be dynamically utilized in the reasoning process, but also the feature data of different fields can be effectively stored and managed, solving the problem of limited model training effect due to the scarcity of annotated data, and improving the cross-domain adaptability and generalization ability of the model; and then by adjusting the data of each target second image data determined by the frequency domain feature alignment operation, the dynamic adjustment of the specific field features can be realized without the need for a large amount of new field annotated data, so that the model can be fine-tuned in the reasoning process to achieve rapid adjustment of model parameters to better adapt to new multi-domain complex scenes, effectively making up for the overfitting problem caused by data imbalance and field-specific training samples in the prior art, better adapting to new multi-domain complex scenes, thereby ensuring that the enhancement process can adapt to the specific needs of different fields. Therefore, based on the method of the embodiment of the present application, through the fusion of memory and decision-making, excellent performance in multi-domain complex scenes is achieved, the accuracy and speed of image enhancement are significantly improved in wide field and high resolution scenes, and the problem of insufficient reasoning effect of image annotation reasoning model in multi-domain complex scenes is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In the attached picture: Figure 1 It is a flowchart of the steps of a multi-domain image intelligent enhancement method based on memory and decision fusion provided in an embodiment of the present application; Figure 2 It is a flowchart of another method for multi-domain image intelligent enhancement based on memory and decision fusion provided by an embodiment of the present application; Figure 3 It is a frequency domain distribution alignment process provided by an embodiment of the present application; Figure 4 This is a complete data flow process provided by the embodiment of the present application; Figure 5 It is a block diagram of a multi-domain image intelligent enhancement device based on memory and decision fusion provided by an embodiment of the present application; Figure 6 is a block diagram of an electronic device according to an embodiment of the present application; Figure 7 It is a block diagram of an electronic device of another embodiment provided by the embodiments of the present application. DETAILED DESCRIPTION

[0012] 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 part of the embodiments of the present application, not all of the embodiments. 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.

[0013] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0014] like Figure 1 As shown, this embodiment provides a multi-domain image intelligent enhancement method based on memory and decision fusion.

[0015] The method may include the following steps: Step 101 , in response to input of first image data, determining a plurality of target second image data from a plurality of second image data stored in a memory unit of a labeled reasoning model.

[0016] Among them, the inference model is used to determine the enhancement data of the first image data based on the second image data stored in the memory unit for the input first image data; the first image data is the image data in the first data field to be entered, and the second image data is the image data in the second data field that has been entered.

[0017] In some embodiments of the present application, in response to the input of the first image data, multiple target second image data are determined from the multiple second image data stored in the memory unit of the labeled reasoning model, in order to dynamically utilize the existing high-quality labeled data in the reasoning process. Through such a process, feature data in different fields can be effectively stored and managed. Specifically, the labeled reasoning model is used to determine the enhanced data of the first image data based on the second image data stored in the memory unit for the input first image data. The first image data is the image data in the first data field to be entered, and the second image data is the image data in the second data field that has been entered. In this way, the problem of limited model training effect due to the scarcity of labeled data can be solved, and the cross-domain adaptability and generalization ability of the model can be improved.

[0018] In a specific example, a medical image analysis system receives a medical image of a new patient (i.e., first image data). The system retrieves multiple annotated medical images (i.e., second image data) by querying the memory unit, which are from different patients and medical scenarios. By utilizing these annotated high-quality image data, the system can generate enhanced medical images of the new patient, which improves image quality and detail recognition capabilities, ensuring more accurate information in subsequent diagnosis.

[0019] Step 102 : determining adjustment data of each target second image data by performing a feature alignment operation from a first frequency domain of the first image data to a target second frequency domain of a plurality of target second image data.

[0020] In some embodiments of the present application, the purpose of performing a feature alignment operation from the first frequency domain of the first image data to the target second frequency domain of multiple target second image data is to achieve dynamic adjustment of specific domain features without the need for a large amount of new domain annotation data. This allows the model to better adapt to new unknown domains. In this process, the first image data and the target second image data are first transformed in the frequency domain to extract their frequency domain features. Then, the frequency domain features of the multiple target second image data are aligned using the frequency domain features of the first image data to determine the adjustment data for each target second image data. The frequency domain refers to the representation of image data in the frequency domain. Through this representation, feature alignment operations can be implemented more efficiently. This ensures that the model can quickly adjust the model parameters through fine-tuning during the inference process to better adapt to new multi-domain complex scenarios.

[0021] In a specific example, an unmanned driving system receives a road condition image (i.e., first image data) from a new environment. The system first performs frequency domain transformations, such as Fourier transform, fast Fourier transform, etc., on the new road condition image and the annotated image (i.e., target second image data) stored in the memory unit to extract their frequency domain features. Then, the system uses the frequency domain features of the new road condition image to align the frequency domain features of multiple target images and determine the adjustment data for each annotated image. In this way, the system can adjust and optimize its model parameters in real time to adapt to the complex road conditions in the new environment, thereby improving the safety and reliability of the unmanned driving system.

[0022] Step 103 , fine-tuning the annotation reasoning model during the reasoning process by determining the adjustment data of the target second image data, thereby determining the enhanced data of the first image data.

[0023] In some embodiments of the present application, the purpose of fine-tuning the annotated reasoning model during the reasoning process by determining the adjustment data of the target second image data is to further optimize the model performance during the reasoning process. First, the target second image data is adjusted by the frequency domain features of the first image data to obtain the adjustment data of the target second image data. Then, these adjustment data are used to fine-tune the annotated reasoning model, that is, to adjust the parameters and feature extraction strategy of the model in real time. The annotated reasoning model is a model that enhances the input image data during the reasoning process. Ultimately, it can ensure that the model parameters can quickly adapt to new domain features, thereby effectively improving the cross-domain generalization ability and enhancement effect of the model.

[0024] In a specific example, an intelligent monitoring system receives a monitoring image (i.e., first image data) from a new environment. The system uses the adjustment data of multiple target second image data stored in the memory unit to fine-tune the annotation reasoning model. In this way, the system can adjust its parameters and feature extraction strategies in real time to adapt to the monitoring needs in the new environment, and finally generate an enhanced monitoring image with improved image clarity and recognition ability, ensuring the monitoring effect in different environments.

[0025] In summary, in the embodiment of the present application, by determining multiple target second image data from multiple second image data stored in the memory unit of the annotated reasoning model, not only can the existing high-quality annotated data be dynamically utilized in the reasoning process, but also the feature data of different fields can be effectively stored and managed, solving the problem of limited model training effect due to the scarcity of annotated data, and improving the cross-domain adaptability and generalization ability of the model; and then by adjusting the data of each target second image data determined by the frequency domain feature alignment operation, the dynamic adjustment of the specific field features can be realized without the need for a large amount of new field annotated data, so that the model can be fine-tuned in the reasoning process to achieve rapid adjustment of model parameters to better adapt to new multi-domain complex scenes, effectively making up for the overfitting problem caused by data imbalance and field-specific training samples in the prior art, better adapting to new multi-domain complex scenes, thereby ensuring that the enhancement process can adapt to the specific needs of different fields. Therefore, based on the method of the embodiment of the present application, through the fusion of memory and decision-making, excellent performance in multi-domain complex scenes is achieved, the accuracy and speed of image enhancement are significantly improved in wide field and high resolution scenes, and the problem of insufficient reasoning effect of image annotation reasoning model in multi-domain complex scenes is solved.

[0026] Figure 2 This is another multi-domain image intelligent enhancement method based on memory and decision fusion provided in an embodiment of the present application.

[0027] The method may include the following steps: Step 201 , in response to input of first image data, determining a plurality of target second image data from a plurality of second image data stored in a memory unit of a labeled reasoning model.

[0028] Among them, the inference model is used to determine the enhancement data of the first image data based on the second image data stored in the memory unit for the input first image data; the first image data is the image data in the first data field to be entered, and the second image data is the image data in the second data field that has been entered.

[0029] The method shown in this step has been explained in step 101 and will not be repeated here.

[0030] Optionally, step 201 includes the following sub-steps: Sub-step 2011, in response to input of first image data, determining a first data capacity of a first image data set where the first image data is located.

[0031] In some embodiments of the present application, in response to the input of the first image data, the first data capacity of the first image data set in which the first image data is located is determined in order to understand the size of the data set so as to perform appropriate processing in subsequent steps. In this process, by analyzing the amount of data in the first image data set, the system can obtain the first data capacity of the data set. The data capacity refers to the number of images in the image data set. This ensures that the system can reasonably allocate resources and adjust the processing strategy according to the size of the data set during the processing, thereby improving the overall processing efficiency and effect.

[0032] In a specific example, an intelligent monitoring system receives a batch of monitoring images (i.e., the first image data set) from a new environment. The system first needs to determine the size of the data set. By analyzing the number of monitoring images in the batch, the system determines that its first data capacity is 500 images. Through this process, the system can adjust resource allocation and processing strategies according to the size of the data set in the subsequent image processing and enhancement process to ensure the efficiency and accuracy of the processing process.

[0033] Sub-step 2012, determining a target data capacity of the target second image data according to the first data capacity and the second data capacity of the memory unit.

[0034] In some embodiments of the present application, the target data capacity of the target second image data is determined based on the first data capacity and the second data capacity of the memory unit, in order to reasonably obtain and use the data in the memory unit in subsequent steps. First, the first data capacity of the first image data set can be determined, and then the target data capacity of the target second image data can be calculated in combination with the total number of second image data stored in the memory unit. Data capacity refers to the number of images in an image data set. This ensures that the system can reasonably allocate resources according to the size of different data sets during processing, thereby improving overall processing efficiency and effectiveness.

[0035] In a specific example, an intelligent monitoring system receives a batch of monitoring images (i.e., the first image data set) from a new environment. The system first determines that the first data capacity of the data set is 500 images. At the same time, the total number of annotated image data (i.e., the second image data) stored in the memory unit is 2,000 images. The system determines through calculation that the target data capacity of the target second image data is 1,000 images, so that the data can be reasonably used in subsequent steps. Through this processing, the system can adjust resource allocation and processing strategies according to the size of the data set in the subsequent image processing and enhancement process to ensure the efficiency and accuracy of the processing process.

[0036] Sub-step 2013, determining a plurality of second image data randomly acquired from the memory unit according to the target data capacity as target second image data.

[0037] In some embodiments of the present application, multiple second image data randomly obtained from the memory unit according to the target data capacity are determined as target second image data, in order to use representative annotated data for enhancement in subsequent processing. In this way, the system can ensure that representative multi-domain image data is obtained from the memory unit, thereby improving the adaptability and generalization ability of the model in different fields. In this process, the system first randomly selects multiple second image data from the memory unit according to the previously determined target data capacity. The random acquisition process is to avoid the bias of data selection and ensure that the acquired target second image data is diverse and representative. The target second image data determined in this way will be used for subsequent image enhancement and processing steps.

[0038] In a specific example, an unmanned driving system receives a road condition image (i.e., first image data) from a new environment. The system randomly selects 1,000 annotated road condition images (i.e., second image data) from the memory unit based on the target data capacity. These road condition images cover different weather, road conditions, and time periods, and are diverse and representative. Through this random acquisition process, the system ensures that the acquired target second image data can be used for subsequent enhancement and processing steps, improving the adaptability and generalization ability of the unmanned driving system in new environments.

[0039] Optionally, sub-step 2013 includes the following sub-steps: Sub-step 20131, randomly acquiring a plurality of target second image data pairs from a plurality of second image data pairs stored in a memory unit according to a target data capacity.

[0040] The second image data pair stores the corresponding second image data and the enhanced data of the second image data in a preset storage format.

[0041] In some embodiments of the present application, multiple target second image data pairs are randomly acquired from multiple second image data pairs stored in a memory unit according to the target data capacity, for the purpose of using representative annotated data for enhancement in subsequent processing. The system can ensure that representative multi-domain image data is acquired from the memory unit to improve the adaptability and generalization ability of the model in different fields. Specifically, first, based on the determined target data capacity, multiple second image data pairs are randomly selected from the memory unit. The second image data pairs store the corresponding second image data and their enhanced data in a preset storage format. The random acquisition process is to avoid bias in data selection and ensure that the acquired target second image data pairs are diverse and representative. In this way, the system can use these data for enhancement and processing in subsequent steps.

[0042] In a specific example, an intelligent monitoring system receives a batch of monitoring images (i.e., first image data) from a new environment. The system randomly selects 500 pairs of labeled monitoring image data pairs (i.e., second image data pairs) from the memory unit according to the target data capacity. These image data pairs store the original image data and the enhanced image data in a preset storage format. Through this random acquisition process, the system ensures that the acquired target second image data pairs are diverse and representative, which is helpful for subsequent image processing and enhancement steps.

[0043] Sub-step 20132, extracting the target second image data from each target second image data pair respectively.

[0044] In some embodiments of the present application, the target second image data is extracted from each target second image data pair respectively, in order to use only necessary image data in subsequent processing steps, thereby improving processing efficiency and effect. According to the previously determined target second image data pair, the system can extract the original target second image data from each image data pair respectively. The target second image data pair stores the corresponding second image data and its enhanced data. Through this extraction process, the system ensures that only necessary image data is obtained for subsequent processing. In this way, the system will obtain the target second image data for subsequent processing.

[0045] In a specific example, an intelligent monitoring system receives a monitoring image (i.e., first image data) of a new environment. The system extracts the original target second image data from each of the 500 pairs of randomly selected and labeled monitoring image data pairs. These extracted image data will be used in subsequent image enhancement and processing steps. Through this extraction process, the system ensures that the necessary image data is obtained, improving processing efficiency and effectiveness.

[0046] Step 202 : determining adjustment data of each target second image data by performing a feature alignment operation from a first frequency domain of the first image data to a target second frequency domain of a plurality of target second image data.

[0047] The method shown in this step has been explained in step 102 and will not be repeated here.

[0048] Optionally, step 202 includes the following sub-steps: Sub-step 2021 , respectively determining the first frequency domain features of the first image data and the target second frequency domain features of each target second image data.

[0049] In some embodiments of the present application, the first frequency domain features of the first image data and the target second frequency domain features of each target second image data are determined respectively, in order to extract important information of the image in the frequency domain so as to perform subsequent feature alignment operations. Specifically, the frequency domain features of the first image data and the target second image data can be extracted by performing frequency domain transformation on the first image data and the target second image data. Frequency domain features refer to the representation of an image in the frequency domain, such as phase spectrum and amplitude spectrum information. In this way, the frequency domain features of the first image data and the target second image data can be obtained, laying the foundation for subsequent feature alignment operations.

[0050] In a specific example, an intelligent monitoring system receives a monitoring image of a new environment (i.e., first image data). The system processes the monitoring image through Fourier transform to extract its frequency domain features. At the same time, the system also performs Fourier transform on the annotated image data (i.e., target second image data) stored in the memory unit. Through this processing, the system obtains the frequency domain features of the first image data and the target second image data, providing the necessary data basis for the subsequent frequency domain feature alignment operation.

[0051] Specifically, Figure 3 As shown, in some embodiments of the present application, sub-step 2021 can be implemented by the following sub-steps: Sub-step 20210, determines the first phase spectrum A1 and the first amplitude spectrum A2 of the first image data, and the target second phase spectrum B1 and the target second amplitude spectrum B2 of the target second image data B based on the Fourier transform operations on the first image data A and the target second image data B, respectively.

[0052] In some embodiments of the present application, the first phase spectrum A1 and the first amplitude spectrum A2 of the first image data A, and the target second phase spectrum B1 and the target second amplitude spectrum B2 of the target second image data B are determined according to the Fourier transform operation on the first image data A and the target second image data B, respectively, in order to extract important information of the image in the frequency domain, so as to perform subsequent feature alignment operations. In this process, the frequency domain features of the first image data A and the target second image data B are extracted by performing Fourier transform. Fourier transform is the process of converting an image from the spatial domain to the frequency domain, and the frequency domain features include phase spectrum and amplitude spectrum information. In this way, the phase spectrum and amplitude spectrum of the first image data A and the target second image data B can be obtained, laying the foundation for the subsequent feature alignment operation.

[0053] In a specific example, an unmanned driving system receives a road condition image (i.e., first image data) from a new environment. The system processes the road condition image through Fourier transform to extract its phase spectrum and amplitude spectrum information. At the same time, the system also performs Fourier transform on the annotated road condition image data (i.e., target second image data) stored in the memory unit to extract its phase spectrum and amplitude spectrum information. Through this processing, the system obtains the frequency domain features of the first image data and the target second image data, providing the necessary data basis for the subsequent frequency domain feature alignment operation.

[0054] Sub-step 2022, performing a distribution alignment operation on each target second frequency domain feature according to the first frequency domain feature, to determine the target second updated frequency domain feature of each target second image data.

[0055] In some embodiments of the present application, the distribution alignment operation of each target second frequency domain feature is performed respectively according to the first frequency domain feature to determine the target second updated frequency domain feature of each target second image data, in order to make the model better adapt to the new domain features through the frequency domain feature alignment operation. Specifically, the frequency domain features of each target second image data are first aligned using the frequency domain features of the first image data. The frequency domain feature alignment operation adjusts the frequency domain information of the target second image data to make it consistent with the frequency domain features of the first image data. The frequency domain feature is the representation of the image in the frequency domain. In this way, the target second updated frequency domain features of each target second image data can be obtained, laying the foundation for the subsequent spatial domain restoration operation.

[0056] In a specific example, an intelligent monitoring system receives a monitoring image (i.e., first image data) of a new environment. The system aligns the frequency domain features of the monitoring image with the frequency domain features of the annotated image data (i.e., target second image data) stored in the memory unit through a frequency domain feature alignment operation. Through this process, the system adjusts the frequency domain features of each target image to obtain updated frequency domain features of each target image. These updated frequency domain features better reflect the features in the new environment and provide the necessary data basis for subsequent image enhancement and processing steps.

[0057] Specifically, Figure 3 As shown, in some embodiments of the present application, the first frequency domain feature includes a first phase spectrum and a first amplitude spectrum, the target second frequency domain feature includes a target second phase spectrum and a target second amplitude spectrum, and sub-step 2022 can be implemented by the following sub-steps: Sub-step 20221, determining the first central area data A0 of A2 in the first amplitude spectrum, and the second central area data B0 of B1 in each second amplitude spectrum.

[0058] In some embodiments of the present application, the first central area data A0 of A2 in the first amplitude spectrum and the second central area data B0 in each second amplitude spectrum B2 are determined in order to realize the alignment operation of the frequency domain features in the subsequent steps, thereby optimizing the image enhancement effect. In this process, the first amplitude spectrum A2 of the first image data A and the second amplitude spectrum B2 of the target second image data B are first extracted by Fourier transform. Then, the system determines the first central area data in the first amplitude spectrum and the second central area data in each second amplitude spectrum. The central area data refers to the frequency information of the central part of the frequency domain image, which usually contains the main features of the image. In this way, the central area frequency information of the two sets of data can be obtained, providing basic data for the subsequent frequency domain feature alignment operation.

[0059] In a specific example, a satellite remote sensing image processing system receives a new terrain image (i.e., first image data). The system extracts the first amplitude spectrum of the terrain image through Fourier transform, and determines the first central area data therefrom. At the same time, the system also performs Fourier transform on the annotated terrain image data (i.e., target second image data) stored in the memory unit, extracts its second amplitude spectrum, and determines the second central area data therefrom. Through this processing, the system obtains the central area frequency information of the first image data and the target second image data, providing the necessary data basis for the subsequent frequency domain feature alignment operation.

[0060] Sub-step 20222, using the first central area data A0 to update the second central area data B0 in each second amplitude spectrum B2 to obtain an updated amplitude spectrum of each target second image data.

[0061] In some embodiments of the present application, the second central area data B0 of B in each second amplitude spectrum is updated with the first central area data A0 to obtain an updated amplitude spectrum of each target second image data B. The purpose is to make the frequency spectrum characteristics of the target image data closer to the distribution characteristics of the unknown field through local adjustment of the frequency domain information. First, the first central area data A0 in the first amplitude spectrum A2 and the target second central area data B0 of B2 in each target second amplitude spectrum are determined. Then, the first central area data A0 replaces the corresponding central area data B0 in each second amplitude spectrum B2. The updated amplitude spectrum better reflects the characteristics of the new field. The central area data is the frequency information of the central part of the frequency domain image, which usually contains the main features of the image. In this way, the system obtains an updated amplitude spectrum for each target second image data, so that it can better adapt to the characteristics of the new field.

[0062] In a specific example, an intelligent monitoring system receives a monitoring image (i.e., first image data) of a new environment. The system extracts the first central area data in the first amplitude spectrum of the monitoring image through Fourier transform, and uses this data to update the second central area data in the second amplitude spectrum of the annotated monitoring image (i.e., target second image data) stored in the memory unit. Through this local adjustment, the system obtains an updated amplitude spectrum for each target monitoring image, and these updated amplitude spectra better reflect the characteristics in the new environment.

[0063] Sub-step 20223, determining the updated amplitude spectrum and the target second phase spectrum corresponding to each target second image data as the target second updated frequency domain features of each target second image data.

[0064] In some embodiments of the present application, the updated amplitude spectrum and the target second phase spectrum corresponding to each target second image data are determined as the target second updated frequency domain features of each target second image data, in order to obtain the complete target second updated frequency domain features by combining the updated amplitude spectrum and phase spectrum, thereby optimizing the image enhancement effect. First, the updated amplitude spectrum and the target second phase spectrum of each target second image data are determined. Then, these two parts of information are combined to determine the target second updated frequency domain features of each target second image data. The updated amplitude spectrum is the amplitude spectrum after the previous adjustment, and the target second phase spectrum is the phase information of the target image. In this way, the system will obtain the complete target second updated frequency domain features, laying the foundation for the subsequent spatial domain restoration operation.

[0065] In a specific example, an intelligent monitoring system receives a monitoring image of a new environment (i.e., first image data). The system obtains the updated amplitude spectrum of the target monitoring image through the previous steps and combines it with the phase spectrum of the image (i.e., the target second phase spectrum). Through this combination, the system determines the target second updated frequency domain features of each target monitoring image.

[0066] Sub-step 2023, determining adjustment data of the second image data of each target according to the spatial domain restoration operation on the second updated frequency domain features of each target.

[0067] In some embodiments of the present application, the adjustment data of the second image data of each target is determined based on the spatial domain restoration operation of the second updated frequency domain features of each target, in order to restore the frequency domain information to the spatial domain, so that the image enhancement effect is more intuitive and practical. First, the second updated frequency domain features of each target are converted back to the spatial domain. The spatial domain is the original representation of the image data. Through this restoration operation, data that is closer to the actual image can be obtained. In this way, the adjustment data of the second image data of each target can be obtained, and these adjustment data retain the high quality characteristics of the original labeled data, while aligning with the unknown domain data in terms of distribution characteristics.

[0068] In a specific example, an unmanned driving system receives a road condition image (i.e., first image data) from a new environment. The system processes the previously updated frequency domain features through a spatial domain restoration operation, such as direct back projection or filtered back projection, and restores these features to the spatial domain to obtain the adjusted data for each target image. Through this restoration operation, the image data generated by the system not only retains the high-quality annotation characteristics, but also better reflects the actual road conditions in the new environment.

[0069] Specifically, Figure 3 As shown, in some embodiments of the present application, sub-step 2023 can be implemented by the following sub-steps: Sub-step 20230, determining the adjustment data C of each target second image data B according to the inverse Fourier transform operation on each target second updated frequency domain feature.

[0070] In some embodiments of the present application, the adjustment data C of each target second image data B is determined according to the inverse Fourier transform operation on each target second updated frequency domain feature, in order to restore the frequency domain information to the spatial domain, so that the image enhancement effect is more intuitive and practical. The inverse Fourier transform is the process of restoring an image from the frequency domain to the spatial domain. Through this restoration operation, data closer to the actual image can be obtained. In this way, the adjustment data C of each target second image data B can be obtained, and these adjustment data retain the high quality characteristics of the original annotated data, while aligning with the unknown domain data in terms of distribution characteristics.

[0071] In a specific example, an unmanned driving system receives a road condition image (i.e., first image data) from a new environment. The system processes the second updated frequency domain features of the target through inverse Fourier transform, restores these features to the spatial domain, and obtains the adjusted data of each target image. Through this restoration operation, the image data generated by the system not only retains the high-quality annotation characteristics, but also better reflects the actual road conditions in the new environment.

[0072] Step 203 , fine-tuning the annotation reasoning model during the reasoning process by determining the adjustment data of the target second image data, thereby determining the enhanced data of the first image data.

[0073] The method shown in this step has been explained in step 103 and will not be repeated here.

[0074] Step 204 , storing the determined enhanced data of the first image data together with the first image data in a memory unit according to a preset storage format, and adding the first data field as a new second data field.

[0075] In some embodiments of the present application, the enhanced data of the determined first image data is stored together with the first image data in a memory unit according to a preset storage format, and the first data field is added as a new second data field, in order to gradually accumulate and enrich domain knowledge in the system. Through this process, the data in the memory unit can be dynamically updated, enabling the system to effectively utilize new domain data in future reasoning and enhancement tasks. The memory unit is a unit that stores labeled data and domain distribution parameters. By storing the enhanced data and the original data together, the consistency and integrity of the data can be ensured. In this way, the system can not only better adapt to the complex needs of multi-domain scenarios, but also gradually improve its cross-domain generalization capabilities.

[0076] In a specific example, a satellite remote sensing image processing system receives a new terrain image (i.e., first image data). The system determines the enhanced data of the image through the previous steps. Then, the system stores the enhanced data and the original image data together in the memory unit according to the preset storage format, and adds the field to which the terrain image belongs as a new second data field. In this way, the system can gradually accumulate more high-quality annotated data of different terrains, improving its adaptability and generalization ability in future remote sensing image processing tasks.

[0077] Optionally, in some cases, considering that the data quality of the first image data is not sufficient for direct subsequent processing, the embodiment of the present application further includes the following additional steps: Step 205: Update the first image data by filtering the first image data.

[0078] In some embodiments of the present application, the first image data is updated by filtering the first image data, in order to improve the quality of the image data, thereby ensuring the accuracy and reliability of subsequent processing. In this process, filtering removes redundant background information, highlights the key feature areas of the foreground, and effectively reduces the interference of background noise on the image enhancement effect. Filtering is a method of image preprocessing, through which the clarity and detail recognition ability of the image can be significantly improved. The first image data updated in this way will have higher quality, making it adaptable to the processing requirements of subsequent steps.

[0079] In a specific example, a satellite remote sensing image processing system receives a low-quality terrain image (i.e., the first image data). The system removes redundant background information and highlights the key feature areas of the terrain by filtering the terrain image. Through this processing, the system significantly reduces the impact of background noise, improves the quality and clarity of the image, and ensures the accuracy and reliability of subsequent image enhancement and processing steps.

[0080] Optionally, considering that in some cases, the data value ranges of the first image data and the second image data are not consistent in the initial case, and the data value range preset for the second image data is known, the embodiment of the present application further includes the following additional steps: Step 206 , performing data mapping processing on the first image data according to the data value range, and updating the first image data.

[0081] In some embodiments of the present application, considering that in some cases, the data value ranges of the first image data and the second image data are not consistent in the initial case, and the data value range preset by the second image data is known, the data mapping processing of the first image data is performed according to the data value range, and the first image data is updated to ensure that the two sets of data are in the same range for subsequent processing. In this process, by using a preset mapping function, the value range of the first image data is adjusted to be consistent with the second image data. This not only helps to ensure the consistency of the data, but also improves the image enhancement effect and the accuracy of the inference model. Data mapping processing refers to the process of converting data from one numerical range to another numerical range. After executing this step, the updated first image data can be more adaptable to subsequent image enhancement and processing steps, thereby improving the effect of the entire system.

[0082] In a specific example, an intelligent monitoring system receives a monitoring image (i.e., first image data) from a new environment, whose value range is 0 to 255, while the value range of the annotated image (i.e., second image data) stored in the memory unit is 0 to 1. The system adjusts the value range of the first image data from 0 to 255 to 0 to 1 through a pre-set mapping function. Through this data mapping process, the system ensures that the two sets of data are in the same value range, which helps the subsequent image enhancement and accurate processing of the inference model.

[0083] In summary, in the embodiment of the present application, by determining multiple target second image data from multiple second image data stored in the memory unit of the annotated reasoning model, not only can the existing high-quality annotated data be dynamically utilized in the reasoning process, but also the feature data of different fields can be effectively stored and managed, solving the problem of limited model training effect due to the scarcity of annotated data, and improving the cross-domain adaptability and generalization ability of the model; and then by adjusting the data of each target second image data determined by the frequency domain feature alignment operation, the dynamic adjustment of the specific field features can be realized without the need for a large amount of new field annotated data, so that the model can be fine-tuned in the reasoning process to achieve rapid adjustment of model parameters to better adapt to new multi-domain complex scenes, effectively making up for the overfitting problem caused by data imbalance and field-specific training samples in the prior art, better adapting to new multi-domain complex scenes, thereby ensuring that the enhancement process can adapt to the specific needs of different fields. Therefore, based on the method of the embodiment of the present application, through the fusion of memory and decision-making, excellent performance in multi-domain complex scenes is achieved, the accuracy and speed of image enhancement are significantly improved in wide field and high resolution scenes, and the problem of insufficient reasoning effect of image annotation reasoning model in multi-domain complex scenes is solved.

[0084] Figure 4The complete data transfer process provided by the embodiment of the present application specifically includes the following steps: Step S1, data preprocessing: The system first receives the input image data and performs necessary preprocessing operations on it. These operations may include normalization, mapping the image pixel values ​​to a standardized range; filtering, removing redundant background information and highlighting the key feature areas of the foreground; and other processing methods such as noise removal. Through these preprocessing operations, the consistency and comparability of the input data are ensured, laying the foundation for subsequent processing.

[0085] Step S2, storing data: After completing data preprocessing, the system stores the processed image data in the memory unit. This step ensures the orderly storage and management of the data, which is convenient for effective call and use in subsequent processing. The stored data includes the normalized image data and its related annotation information.

[0086] Step S3, in the memory unit, the system extracts relevant information from the stored data when needed. The memory unit not only stores the existing high-quality annotated data, but also dynamically updates and accumulates new domain feature data. Through this operation, the system can continuously enrich its domain knowledge and improve its adaptability and generalization ability in multi-domain scenarios.

[0087] Step S4: Data extraction (including multiple sub-steps): Sub-step S41, data extraction: the system extracts the annotated image data related to the current task from the memory unit. By querying the memory unit, the system obtains a plurality of stored second image data, which contain the system's high-quality annotated data in different fields; Sub-step S42, Fourier transform: The system performs Fourier transform on the first image data and the target second image data respectively to extract their frequency domain features, including phase spectrum and amplitude spectrum information. Fourier transform is the process of converting an image from the spatial domain to the frequency domain, and this conversion can more efficiently perform feature extraction and alignment operations; Sub-step S43, frequency domain distribution alignment: using the frequency domain features of the first image data, align the frequency domain features of the target second image data. Specifically, the system uses the frequency domain information of the central area of ​​the first image data to update the frequency domain information of the corresponding area of ​​the target second image data, so that the frequency domain features of the target second image data are closer to the distribution features of the unknown area; Sub-step S44, inverse Fourier transform: The system performs inverse Fourier transform on the target second image data after the frequency domain distribution alignment processing, and restores the frequency domain information to the spatial domain. This step converts the adjusted frequency domain features back to the spatial domain representation of the image, thereby generating adjusted image data, ready for subsequent image enhancement and processing.

[0088] Step S5, fusion optimization: The system fuses and optimizes the extracted adjustment data. This step combines data from different fields to optimize the model's parameters and feature extraction strategies to ensure that the system can adapt to new unknown fields.

[0089] Ultimately, the optimized data was used in actual application scenarios, such as target detection and image recognition, improving the system's performance in complex environments in multiple fields.

[0090] refer to Figure 5 , which shows a multi-domain image intelligent enhancement device 30 based on memory and decision fusion provided by an embodiment of the present application, including: The reference selection module 301 is used to determine a plurality of target second image data from a plurality of second image data stored in a memory unit of the annotated reasoning model in response to input of the first image data; the reasoning model is used to determine the enhancement data of the first image data for the input first image data according to the second image data stored in the memory unit; the first image data is the image data under the first data domain to be entered, and the second image data is the image data under the second data domain that has been entered; An alignment adjustment module 302 is used to determine adjustment data of each target second image data by performing a feature alignment operation of a first frequency domain of the first image data to a target second frequency domain of a plurality of target second image data; The reasoning enhancement module 303 is used to determine the enhancement data of the first image data by fine-tuning the annotation reasoning model during the reasoning process through the determined adjustment data of the target second image data.

[0091] Optionally, the multi-domain image intelligent enhancement device 30 based on memory and decision fusion further includes: The filtering update module is used to update the first image data by filtering the first image data.

[0092] Optionally, the second image data has a preset data value range, and the multi-domain image intelligent enhancement device 30 based on memory and decision fusion further includes: The mapping update module is used to process the data mapping of the first image data according to the data value range and update the first image data.

[0093] Optionally, the reference selection module 301 includes: a capacity confirmation submodule, configured to determine, in response to input of the first image data, a first data capacity of a first image data set where the first image data is located; a quantity matching submodule, used to determine a target data capacity of the target second image data according to the first data capacity and the second data capacity of the memory unit; The target extraction submodule is used to determine a plurality of second image data randomly acquired from the memory unit according to the target data capacity as target second image data.

[0094] Optionally, the target extraction submodule includes: A data pair extraction unit, configured to randomly obtain a plurality of target second image data pairs from a plurality of second image data pairs stored in a memory unit according to a target data capacity; the second image data pairs store corresponding second image data and enhanced data of the second image data in a preset storage format; The image extraction unit is used to extract the target second image data from each target second image data pair.

[0095] Optionally, the alignment adjustment module 302 includes: The frequency domain feature extraction submodule is used to respectively determine the first frequency domain feature of the first image data and the target second frequency domain feature of each target second image data.

[0096] Optionally, the frequency domain feature extraction submodule includes: The Fourier transform unit is used to determine a first phase spectrum and a first amplitude spectrum of the first image data, and a target second phase spectrum and a target second amplitude spectrum of the target second image data according to Fourier transform operations on the first image data and the target second image data respectively.

[0097] The distribution alignment submodule is used to perform distribution alignment operations on each target second frequency domain feature according to the first frequency domain feature, so as to determine the target second updated frequency domain feature of each target second image data.

[0098] Optionally, the first frequency domain feature includes a first phase spectrum and a first amplitude spectrum, the target second frequency domain feature includes a target second phase spectrum and a target second amplitude spectrum, and the distribution alignment submodule includes: A center confirmation unit, used to determine first center area data in the first amplitude spectrum, and second center area data in each second amplitude spectrum; a center replacement unit, for updating the second center area data in each second amplitude spectrum with the first center area data to obtain an updated amplitude spectrum of each target second image data; The frequency domain updating unit is used to determine the updated amplitude spectrum and the target second phase spectrum corresponding to each target second image data as the target second updated frequency domain features of each target second image data.

[0099] The spatial domain restoration submodule is used to determine the adjustment data of the second image data of each target according to the spatial domain restoration operation on the second updated frequency domain features of each target.

[0100] Optionally, the spatial domain restoration submodule includes: The inverse Fourier transform unit is used to determine the adjustment data of each target second image data according to the inverse Fourier transform operation on each target second updated frequency domain feature.

[0101] Optionally, the multi-domain image intelligent enhancement device 30 based on memory and decision fusion further includes: The memory unit updating module is used to store the determined enhanced data of the first image data together with the first image data in the memory unit according to a preset storage format, and add the first data field as a new second data field.

[0102] In summary, in the embodiment of the present application, by determining multiple target second image data from multiple second image data stored in the memory unit of the annotated reasoning model, not only can the existing high-quality annotated data be dynamically utilized in the reasoning process, but also the feature data of different fields can be effectively stored and managed, solving the problem of limited model training effect due to the scarcity of annotated data, and improving the cross-domain adaptability and generalization ability of the model; and then by adjusting the data of each target second image data determined by the frequency domain feature alignment operation, the dynamic adjustment of the specific field features can be realized without the need for a large amount of new field annotated data, so that the model can be fine-tuned in the reasoning process to achieve rapid adjustment of model parameters to better adapt to new multi-domain complex scenes, effectively making up for the overfitting problem caused by data imbalance and field-specific training samples in the prior art, better adapting to new multi-domain complex scenes, thereby ensuring that the enhancement process can adapt to the specific needs of different fields. Therefore, based on the method of the embodiment of the present application, through the fusion of memory and decision-making, excellent performance in multi-domain complex scenes is achieved, the accuracy and speed of image enhancement are significantly improved in wide field and high resolution scenes, and the problem of insufficient reasoning effect of image annotation reasoning model in multi-domain complex scenes is solved.

[0103] Reference Figure 6 , the electronic device 500 may include one or more of the following components: a processing component 502 , a memory 504 , a power component 506 , a multimedia component 508 , an audio component 510 , an input / output (I / O) interface 512 , a sensor component 514 , and a communication component 516 .

[0104] The processing component 502 generally controls the overall operation of the electronic device 500, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 502 may include one or more processors 520 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 502 may include one or more modules to facilitate the interaction between the processing component 502 and other components. For example, the processing component 502 may include a multimedia module to facilitate the interaction between the multimedia component 508 and the processing component 502.

[0105] The memory 504 is used to store various types of data to support the operation of the electronic device 500. Examples of such data include instructions for any application or method operating on the electronic device 500, contact data, phone book data, messages, pictures, multimedia, etc. The memory 504 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0106] The power supply component 506 provides power to the various components of the electronic device 500. The power supply component 506 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 500.

[0107] The multimedia component 508 includes an interface that provides an output interface between the electronic device 500 and the user. In some embodiments, the interface may include a liquid crystal display (LCD) and a touch panel (TP). If the interface includes a touch panel, the interface may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 508 includes a front camera and / or a rear camera. When the electronic device 500 is in an operating mode, such as a shooting mode or a multimedia mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0108] The audio component 510 is used to output and / or input audio signals. For example, the audio component 510 includes a microphone (MIC), and when the electronic device 500 is in an operation mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is used to receive an external audio signal. The received audio signal can be further stored in the memory 504 or sent via the communication component 516. In some embodiments, the audio component 510 also includes a speaker for outputting audio signals.

[0109] The input / output I / O interface 512 provides an interface between the processing component 502 and the peripheral interface module, which may be a keyboard, a click wheel, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0110] The sensor assembly 514 includes one or more sensors for providing various aspects of status assessment for the electronic device 500. For example, the sensor assembly 514 can detect the open / closed state of the electronic device 500, the relative positioning of components, such as the display and keypad of the electronic device 500, and the sensor assembly 514 can also detect the position change of the electronic device 500 or a component of the electronic device 500, the presence or absence of user contact with the electronic device 500, the orientation or acceleration / deceleration of the electronic device 500, and the temperature change of the electronic device 500. The sensor assembly 514 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 514 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 514 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0111] The communication component 516 is used to facilitate wired or wireless communication between the electronic device 500 and other devices. The electronic device 500 can access a wireless network based on a communication standard, such as WiFi, a carrier network (such as 2G, 3G, 4G or 5G), or a combination thereof. In an exemplary embodiment, the communication component 516 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 516 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0112] In an exemplary embodiment, the electronic device 500 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to implement the methods provided in the embodiments of the present application.

[0113] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 504 including instructions, and the instructions can be executed by a processor 520 of an electronic device 500 to perform the above method. For example, the non-transitory storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0114] Figure 7 FIG. 6 is a block diagram of an electronic device 600 according to another embodiment of the present invention. For example, the electronic device 600 may be provided as a server.

[0115] Reference Figure 7 , the electronic device 600 includes a processing component 622, which further includes one or more processors, and a memory resource represented by a memory 632, for storing instructions that can be executed by the processing component 622, such as an application. The application stored in the memory 632 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 622 is configured to execute instructions to perform the method provided in the embodiment of the present application.

[0116] The electronic device 600 may also include a power supply component 626 configured to perform power management of the electronic device 600, a wired or wireless network interface 650 configured to connect the electronic device 600 to a network, and an input / output (I / O) interface 658. The electronic device 600 may operate based on an operating system stored in the memory 632, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™ or the like.

[0117] It should be noted that, for the sake of simplicity, the method embodiments of the present application are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described action sequence, because according to the embodiments of the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0118] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the application disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0119] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A multi-domain image intelligent enhancement method based on memory and decision fusion, characterized in that: include: In response to input of the first image data, determining a plurality of target second image data from a plurality of second image data stored in a memory unit of the annotation reasoning model; The inference model is used to determine the enhancement data of the first image data according to the second image data stored in the memory unit for the input first image data; The first image data is image data in a first data field to be entered, and the second image data is image data in a second data field that has been entered; Determine adjustment data of each of the target second image data by performing a feature alignment operation from a first frequency domain of the first image data to a target second frequency domain of a plurality of the target second image data; The enhanced data of the first image data is determined by fine-tuning the labeled reasoning model during the reasoning process using the determined adjustment data of the target second image data.

2. The multi-domain image intelligent enhancement method based on memory and decision fusion as claimed in claim 1, characterized in that: The second image data has a preset data value range, and the multi-domain image intelligent enhancement method based on memory and decision fusion also includes: The first image data is updated by filtering the first image data and / or performing data mapping processing on the first image data according to the data value range.

3. The multi-domain image intelligent enhancement method based on memory and decision fusion as claimed in claim 1, characterized in that: The method of determining a plurality of target second image data from a plurality of second image data stored in a memory unit of the annotation reasoning model in response to the input of the first image data comprises: In response to input of the first image data, determining a first data capacity of a first image data set where the first image data is located; determining a target data capacity of the target second image data according to the first data capacity and the second data capacity of the memory unit; A plurality of second image data randomly acquired from the memory unit according to the target data capacity are determined as the target second image data.

4. The multi-domain image intelligent enhancement method based on memory and decision fusion as claimed in claim 3, characterized in that: The step of determining the plurality of second image data randomly acquired from the memory unit according to the target data capacity as the target second image data comprises: A plurality of target second image data pairs are randomly acquired from a plurality of second image data pairs stored in the memory unit according to the target data capacity; the second image data pairs store corresponding second image data and enhanced data of the second image data in a preset storage format; The target second image data is extracted respectively from each of the target second image data pairs.

5. The multi-domain image intelligent enhancement method based on memory and decision fusion as claimed in claim 1, characterized in that: The step of determining the adjustment data of each of the target second image data by performing a feature alignment operation from the first frequency domain of the first image data to the target second frequency domain of the plurality of target second image data comprises: respectively determining a first frequency domain feature of the first image data and a target second frequency domain feature of each of the target second image data; Performing a distribution alignment operation on each of the target second frequency domain features according to the first frequency domain features, respectively, to determine a target second updated frequency domain feature of each of the target second image data; According to the spatial domain restoration operation on each of the target second updated frequency domain features, the adjustment data of each of the target second image data is determined.

6. The multi-domain image intelligent enhancement method based on memory and decision fusion as claimed in claim 5, characterized in that: The first frequency domain feature includes a first phase spectrum and a first amplitude spectrum, the target second frequency domain feature includes a target second phase spectrum and a target second amplitude spectrum, and the distribution alignment operation is respectively performed on each of the target second frequency domain features according to the first frequency domain feature to determine the target second updated frequency domain feature of each of the target second image data, including: Determine first central area data in the first amplitude spectrum, and second central area data in each of the second amplitude spectra; Updating the second central area data in each of the second amplitude spectrums with the first central area data to obtain an updated amplitude spectrum of each of the target second image data; An updated amplitude spectrum and a target second phase spectrum corresponding to each of the target second image data are determined as a target second updated frequency domain feature of each of the target second image data.

7. The multi-domain image intelligent enhancement method based on memory and decision fusion as claimed in claim 1, characterized in that: The multi-domain image intelligent enhancement method based on memory and decision fusion also includes: The determined enhancement data of the first image data is stored together with the first image data in the memory unit according to a preset storage format, and the first data field is added as a new second data field.

8. A multi-domain image intelligent enhancement device based on memory and decision fusion, characterized in that: include: A reference selection module is used to determine a plurality of target second image data from a plurality of second image data stored in a memory unit of the annotation reasoning model in response to input of the first image data; the reasoning model is used to determine the enhancement data of the first image data based on the second image data stored in the memory unit for the input first image data; The first image data is image data in a first data field to be entered, and the second image data is image data in a second data field that has been entered; an alignment adjustment module, configured to determine adjustment data of each of the target second image data by performing a feature alignment operation from a first frequency domain of the first image data to a target second frequency domain of a plurality of the target second image data; The reasoning enhancement module is used to determine the enhancement data of the first image data by fine-tuning the labeled reasoning model during the reasoning process through the determined adjustment data of the target second image data.

9. An electronic device, characterized in that: include: A processor, a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the steps of the multi-domain image intelligent enhancement method based on memory and decision fusion as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the steps of the multi-domain image intelligent enhancement method based on memory and decision fusion as described in any one of claims 1 to 7.

Images