An image scene association optimization processing method based on AI

Abstract

The present invention relates to the field of image scene association optimization, and in particular to an image scene association optimization processing method based on AI, comprising: obtaining image scene association features using real-time image data; establishing an image scene association optimization model based on AI according to the image scene association features using the real-time image data; obtaining image scene association optimization processing results using the image scene association optimization model, establishing a pixel-level image recognition improvement model based on AI technology, and being able to automatically identify key features in an image, comprehensively considering factors unexpected to the image content, such as equipment and signals, to accurately and efficiently establish an image optimization solution.

Description

Technical Field

[0001] The present invention relates to the field of image scene association optimization, and in particular to an image scene association optimization processing method based on AI. Background Art

[0002] In traditional image processing schemes, the image area that needs to be optimized is processed as a whole. Excessively blurred regional processing will cause the final output image to be distorted, with a large error from the original image or target expectation. At the same time, the blurring technology does not retain enough details, which may cause detail loss and cause the image style to be far from the original. Simple optimization schemes have limited ability to adapt to different contents, require large computing resources, and are not efficient. When the optimization results are finally output, it is also easy to amplify image noise while reducing noise. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention provides an image scene association optimization processing method based on AI, which establishes a pixel-level optimization model of the image through AI technology, and optimizes each pixel of the image to improve the optimization capability and the integrity of the final image.

[0004] To achieve the above object, the present invention provides an AI-based image scene association optimization processing method, comprising:

[0005] S1, using real-time image data to obtain image scene correlation features;

[0006] S2. Using the real-time image data to establish an image scene association optimization model based on AI according to image scene association features;

[0007] S3. Obtaining an image scene association optimization processing result by using the image scene association optimization model.

[0008] Preferably, the method of obtaining image scene association features using real-time image data includes:

[0009] S1-1, collect real-time image data;

[0010] S1-2, establishing static features and dynamic features of the real-time image data respectively according to the real-time image data;

[0011] S1-3. Utilize the static features and dynamic features of the real-time image data as image scene association features.

[0012] Furthermore, establishing static features and dynamic features of the real-time image data respectively according to the real-time image data includes:

[0013] S1-2-1, collecting real-time image data to obtain image resolution and image target type corresponding to the real-time image data as static features of the real-time image data;

[0014] S1-2-2, acquiring image exposure and image source signal according to the corresponding acquisition device of the real-time image data;

[0015] S1-2-3, using the real-time image data to obtain image texture features;

[0016] S1-2-4. Utilize the image exposure, image source signal and image texture features as dynamic features of real-time image data.

[0017] Furthermore, using the real-time image data to establish an image scene association optimization model based on AI according to the image scene association features includes:

[0018] S2-1, using the real-time image data to establish an image target area recognition model based on AI according to the image scene association features corresponding to the dynamic features;

[0019] S2-2, using the image scene association features to establish a basic image scene association optimization model based on AI;

[0020] S2-3. Utilize the image target region recognition model and the basic image scene association optimization model as the image scene association optimization model.

[0021] Furthermore, using the real-time image data to establish an image target area recognition model based on AI according to the image scene associated features corresponding to the dynamic features includes:

[0022] S2-1-1, using the real-time image data to establish a rough outline of an image target area of ​​the real-time image data according to image texture features corresponding to dynamic features of image scene association features;

[0023] S2-1-2, obtaining a rough outline of an image target area corresponding to the historical image data according to an image source signal of a dynamic feature corresponding to an image scene association feature of the real-time image data;

[0024] S2-1-3, using the rough outline of the image target area of ​​the historical image data to divide into a first training set;

[0025] S2-1-4, using the rough outline of the image target area of ​​the real-time image data to divide into a first verification set;

[0026] S2-1-5, using the first training set as input and the image area corresponding to the first training set as output, and establishing a basic image target area recognition model through AI-based training;

[0027] S2-1-6, using the first verification set to input the basic image target area recognition model to perform model verification processing to obtain a basic image target area recognition verification output;

[0028] S2-1-7, determine whether the basic image target area recognition verification output corresponds to the image texture feature of the real-time image data according to the image scene association feature corresponding to the dynamic feature, if so, execute S2-1-8. Otherwise, obtain the rough outline of the image target area of ​​the historical image data according to the image texture feature of the real-time image data according to the image scene association feature corresponding to the dynamic feature, and update the first training set, and return to S2-1-5;

[0029] S2-1-8. Determine whether the image exposure corresponding to the basic image target area recognition verification output is consistent with the image exposure of the dynamic features corresponding to the image scene associated features of the real-time image data. If so, use the basic image target area recognition model as the image target area recognition model. Otherwise, obtain the rough outline of the image target area of ​​the historical image data according to the image exposure of the dynamic features corresponding to the image scene associated features of the real-time image data, update the first training set, and return to S2-1-5.

[0030] Furthermore, using the image scene association features to establish a basic image scene association optimization model based on AI includes:

[0031] S2-2-1, using the dynamic characteristics of the image scene association characteristics to correspond to the signal frequency of the image source signal to respectively obtain historical normal image data and historical abnormal image data;

[0032] S2-2-2, obtaining abnormal pixels corresponding to the historical abnormal image data;

[0033] S2-2-3, obtaining the image exposure, image source signal and image texture features corresponding to the dynamic features of the historical abnormal image data to respectively establish a first data set, a second data set and a third data set;

[0034] S2-2-4. Using the first data set as input and the abnormal pixels corresponding to the first data set as output, an abnormal exposure recognition model is established through training based on AI;

[0035] S2-2-5. Using the second data set as input and the abnormal pixels corresponding to the second data set as output, an abnormal source signal recognition model is established through training based on AI;

[0036] S2-2-6. Using the third data set as input and the abnormal pixels corresponding to the third data set as output, an abnormal texture feature recognition model is established through training based on AI;

[0037] S2-2-7. Use the abnormal exposure recognition model, abnormal source signal recognition model and abnormal texture feature recognition model to perform model output verification processing based on historical normal image data to obtain a basic image scene association optimization model.

[0038] Furthermore, the abnormal exposure recognition model, the abnormal source signal recognition model and the abnormal texture feature recognition model are used to perform model output verification processing according to historical normal image data to obtain a basic image scene association optimization model, including:

[0039] S2-2-7-1, according to the dynamic characteristics of the image scene associated characteristics corresponding to the historical normal image data, respectively obtain the image exposure, image source signal and image texture characteristics to establish an exposure verification set, a source signal verification set and a texture feature verification set in turn;

[0040] S2-2-7-2, using the exposure verification set to input the abnormal exposure recognition model to obtain the model output of the abnormal exposure recognition model;

[0041] S2-2-7-3, using the source signal verification set to input the abnormal source signal recognition model to obtain the model output of the abnormal source signal recognition model;

[0042] S2-2-7-4, using the texture feature verification set to input the abnormal texture feature recognition model to obtain the model output of the abnormal texture feature recognition model;

[0043] S2-2-7-5, determine whether the model output of the abnormal exposure recognition model exists, if so, use the exposure verification set to update the first data set and return to S2-2-4, otherwise, execute S2-2-7-6;

[0044] S2-2-7-6, determine whether the model output of the abnormal source signal identification model exists, if so, use the source signal verification set to update the second data set and return to S2-2-5, otherwise, execute S2-2-7-7;

[0045] S2-2-7-7. Determine whether the model output of the abnormal texture feature recognition model exists. If so, use the texture feature verification set to update the third data set and return to S2-2-6. Otherwise, use the abnormal exposure recognition model, abnormal source signal recognition model and abnormal texture feature recognition model as the basic image scene association optimization model.

[0046] Furthermore, obtaining the image scene association optimization processing result by using the image scene association optimization model includes:

[0047] S3-1, using the image target area recognition model of the real-time image data input image scene association optimization model to respectively obtain the image target area of ​​the real-time image data and the image non-target area of ​​the real-time image data;

[0048] S3-2, using the dynamic features of the image scene association features corresponding to the real-time image data to input the basic image scene association optimization model of the image scene association optimization model to obtain abnormal exposure recognition results, abnormal source signal recognition results and abnormal texture feature recognition results respectively;

[0049] S3-3, using the image target area and the image non-target area of ​​the real-time image data to perform restoration verification processing according to the static features of the image scene associated features to obtain a restoration verification result of the real-time image data;

[0050] S3-4. Utilizing the restoration verification result of the real-time image data, the image scene association optimization processing result is obtained according to the abnormal exposure recognition result, the abnormal source signal recognition result and the abnormal texture feature recognition result.

[0051] Furthermore, the restoration verification result of the real-time image data obtained by performing restoration verification processing on the image target area and the image non-target area of ​​the real-time image data according to the static features of the image scene associated features includes:

[0052] S3-3-1, determine whether the image target area, image non-target area and static features of the image scene associated features of the real-time image data correspond to the same image target type, if so, execute S3-3-2, otherwise, return to S1-2-1;

[0053] S3-3-2. Determine whether the image resolution corresponding to the image target area of ​​the real-time image data, the image resolution corresponding to the image non-target area, and the image resolution corresponding to the static features of the image scene associated features are all consistent. If so, the restoration verification result of the real-time image data is passed. Otherwise, the restoration verification result of the real-time image data is failed, and the real-time image data is updated and returned to S1-1.

[0054] Furthermore, the image scene association optimization processing results are obtained by using the restoration verification results of the real-time image data according to the abnormal exposure recognition results, the abnormal source signal recognition results and the abnormal texture feature recognition results, including:

[0055] S3-4-1, determine whether the number of restoration verification processes corresponding to the restoration verification result of the real-time image data is greater than 3, if so, return to S2-1-1, otherwise, execute S3-4-2;

[0056] S3-4-2, determine whether the abnormal exposure recognition result exists, if so, use the image exposure of the dynamic feature corresponding to the historical normal image data to adjust the real-time image data according to the abnormal exposure recognition result, and return to S3-2, otherwise, execute S3-4-3;

[0057] S3-4-3, determine whether the abnormal source signal identification result exists, if so, use the image source signal of the dynamic feature corresponding to the historical normal image data to perform noise reduction processing on the image source signal of the dynamic feature corresponding to the real-time image data, and return to S3-2, otherwise, execute S3-4-4;

[0058] S3-4-4, determine whether the abnormal texture feature recognition result exists, if so, use the image texture features of the dynamic features corresponding to the historical normal image data to update the real-time image data according to the abnormal texture feature recognition result, and return to S3-2, otherwise, execute S3-4-5;

[0059] S3-4-5. Determine whether the real-time image data at the current moment and the image target area and image non-target area corresponding to the restoration verification result are consistent. If so, use the real-time image data at the current moment as the image scene association optimization processing result. Otherwise, use the real-time image data at the current moment for update processing and return to S1-2.

[0060] Compared with the closest prior art, the present invention has the following beneficial effects:

[0061] Based on AI technology, a pixel-level image recognition improvement model is established. The model deeply explores the potential of each pixel and aims to improve the overall quality of the image through meticulous adjustments while maintaining the integrity and authenticity of the image content. It can also automatically identify key features in the image, comprehensively consider factors other than image content, such as equipment and signals, and accurately and efficiently establish image optimization solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flow chart of an AI-based image scene association optimization processing method provided by the present invention. DETAILED DESCRIPTION

[0063] The specific implementation modes of the present invention will be further described in detail below in conjunction with the accompanying drawings.

[0064] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Embodiment 1:

[0066] The present invention provides an image scene association optimization processing method based on AI, such as Figure 1 As shown, including:

[0067] S1, using real-time image data to obtain image scene correlation features;

[0068] S2. Using the real-time image data to establish an image scene association optimization model based on AI according to image scene association features;

[0069] S3. Obtaining an image scene association optimization processing result by using the image scene association optimization model.

[0070] S1 specifically includes:

[0071] S1-1, collect real-time image data;

[0072] S1-2, establishing static features and dynamic features of the real-time image data respectively according to the real-time image data;

[0073] S1-3. Utilize the static features and dynamic features of the real-time image data as image scene association features.

[0074] S1-2 specifically includes:

[0075] S1-2-1, collecting real-time image data to obtain image resolution and image target type corresponding to the real-time image data as static features of the real-time image data;

[0076] S1-2-2, acquiring image exposure and image source signal according to the corresponding acquisition device of the real-time image data;

[0077] S1-2-3, using the real-time image data to obtain image texture features;

[0078] S1-2-4. Utilize the image exposure, image source signal and image texture features as dynamic features of real-time image data.

[0079] S2 specifically includes:

[0080] S2-1, using the real-time image data to establish an image target area recognition model based on AI according to the image scene association features corresponding to the dynamic features;

[0081] S2-2, using the image scene association features to establish a basic image scene association optimization model based on AI;

[0082] S2-3. Utilize the image target region recognition model and the basic image scene association optimization model as the image scene association optimization model.

[0083] S2-1 specifically includes:

[0084] S2-1-1, using the real-time image data to establish a rough outline of an image target area of ​​the real-time image data according to image texture features corresponding to dynamic features of image scene association features;

[0085] S2-1-2, obtaining a rough outline of an image target area corresponding to the historical image data according to an image source signal of a dynamic feature corresponding to an image scene association feature of the real-time image data;

[0086] S2-1-3, using the rough outline of the image target area of ​​the historical image data to divide into a first training set;

[0087] S2-1-4, using the rough outline of the image target area of ​​the real-time image data to divide into a first verification set;

[0088] S2-1-5, using the first training set as input and the image area corresponding to the first training set as output, and establishing a basic image target area recognition model through AI-based training;

[0089] S2-1-6, using the first verification set to input the basic image target area recognition model to perform model verification processing to obtain a basic image target area recognition verification output;

[0090] S2-1-7, determine whether the basic image target area recognition verification output corresponds to the image texture feature of the real-time image data according to the image scene association feature corresponding to the dynamic feature, if so, execute S2-1-8. Otherwise, obtain the rough outline of the image target area of ​​the historical image data according to the image texture feature of the real-time image data according to the image scene association feature corresponding to the dynamic feature, and update the first training set, and return to S2-1-5;

[0091] S2-1-8. Determine whether the image exposure corresponding to the basic image target area recognition verification output is consistent with the image exposure of the dynamic features corresponding to the image scene associated features of the real-time image data. If so, use the basic image target area recognition model as the image target area recognition model. Otherwise, obtain the rough outline of the image target area of ​​the historical image data according to the image exposure of the dynamic features corresponding to the image scene associated features of the real-time image data, update the first training set, and return to S2-1-5.

[0092] S2-2 specifically includes:

[0093] S2-2-1, using the dynamic characteristics of the image scene association characteristics to correspond to the signal frequency of the image source signal to respectively obtain historical normal image data and historical abnormal image data;

[0094] S2-2-2, obtaining abnormal pixels corresponding to the historical abnormal image data;

[0095] S2-2-3, obtaining the image exposure, image source signal and image texture features corresponding to the dynamic features of the historical abnormal image data to respectively establish a first data set, a second data set and a third data set;

[0096] S2-2-4. Using the first data set as input and the abnormal pixels corresponding to the first data set as output, an abnormal exposure recognition model is established through training based on AI;

[0097] S2-2-5. Using the second data set as input and the abnormal pixels corresponding to the second data set as output, an abnormal source signal recognition model is established through training based on AI;

[0098] S2-2-6. Using the third data set as input and the abnormal pixels corresponding to the third data set as output, an abnormal texture feature recognition model is established through training based on AI;

[0099] S2-2-7. Use the abnormal exposure recognition model, abnormal source signal recognition model and abnormal texture feature recognition model to perform model output verification processing based on historical normal image data to obtain a basic image scene association optimization model.

[0100] S2-2-7 specifically includes:

[0101] S2-2-7-1, according to the dynamic characteristics of the image scene associated characteristics corresponding to the historical normal image data, respectively obtain the image exposure, image source signal and image texture characteristics to establish an exposure verification set, a source signal verification set and a texture feature verification set in turn;

[0102] S2-2-7-2, using the exposure verification set to input the abnormal exposure recognition model to obtain the model output of the abnormal exposure recognition model;

[0103] S2-2-7-3, using the source signal verification set to input the abnormal source signal recognition model to obtain the model output of the abnormal source signal recognition model;

[0104] S2-2-7-4, using the texture feature verification set to input the abnormal texture feature recognition model to obtain the model output of the abnormal texture feature recognition model;

[0105] S2-2-7-5, determine whether the model output of the abnormal exposure recognition model exists, if so, use the exposure verification set to update the first data set and return to S2-2-4, otherwise, execute S2-2-7-6;

[0106] S2-2-7-6, determine whether the model output of the abnormal source signal identification model exists, if so, use the source signal verification set to update the second data set and return to S2-2-5, otherwise, execute S2-2-7-7;

[0107] S2-2-7-7. Determine whether the model output of the abnormal texture feature recognition model exists. If so, use the texture feature verification set to update the third data set and return to S2-2-6. Otherwise, use the abnormal exposure recognition model, abnormal source signal recognition model and abnormal texture feature recognition model as the basic image scene association optimization model.

[0108] S3 specifically includes:

[0109] S3-1, using the image target area recognition model of the real-time image data input image scene association optimization model to respectively obtain the image target area of ​​the real-time image data and the image non-target area of ​​the real-time image data;

[0110] S3-2, using the dynamic features of the image scene association features corresponding to the real-time image data to input the basic image scene association optimization model of the image scene association optimization model to obtain abnormal exposure recognition results, abnormal source signal recognition results and abnormal texture feature recognition results respectively;

[0111] S3-3, using the image target area and the image non-target area of ​​the real-time image data to perform restoration verification processing according to the static features of the image scene associated features to obtain a restoration verification result of the real-time image data;

[0112] S3-4. Utilizing the restoration verification result of the real-time image data, the image scene association optimization processing result is obtained according to the abnormal exposure recognition result, the abnormal source signal recognition result and the abnormal texture feature recognition result.

[0113] S3-3 specifically includes:

[0114] S3-3-1, determine whether the image target area, image non-target area and static features of the image scene associated features of the real-time image data correspond to the same image target type, if so, execute S3-3-2, otherwise, return to S1-2-1;

[0115] S3-3-2. Determine whether the image resolution corresponding to the image target area of ​​the real-time image data, the image resolution corresponding to the image non-target area, and the image resolution corresponding to the static features of the image scene associated features are all consistent. If so, the restoration verification result of the real-time image data is passed. Otherwise, the restoration verification result of the real-time image data is failed, and the real-time image data is updated and returned to S1-1.

[0116] S3-4 specifically includes:

[0117] S3-4-1, determine whether the number of restoration verification processes corresponding to the restoration verification result of the real-time image data is greater than 3, if so, return to S2-1-1, otherwise, execute S3-4-2;

[0118] S3-4-2, determine whether the abnormal exposure recognition result exists, if so, use the image exposure of the dynamic feature corresponding to the historical normal image data to adjust the real-time image data according to the abnormal exposure recognition result, and return to S3-2, otherwise, execute S3-4-3;

[0119] S3-4-3, determine whether the abnormal source signal identification result exists, if so, use the image source signal of the dynamic feature corresponding to the historical normal image data to perform noise reduction processing on the image source signal of the dynamic feature corresponding to the real-time image data, and return to S3-2, otherwise, execute S3-4-4;

[0120] S3-4-4, determine whether the abnormal texture feature recognition result exists, if so, use the image texture features of the dynamic features corresponding to the historical normal image data to update the real-time image data according to the abnormal texture feature recognition result, and return to S3-2, otherwise, execute S3-4-5;

[0121] S3-4-5. Determine whether the real-time image data at the current moment and the image target area and image non-target area corresponding to the restoration verification result are consistent. If so, use the real-time image data at the current moment as the image scene association optimization processing result. Otherwise, use the real-time image data at the current moment for update processing and return to S1-2.

[0122] In the present embodiment, an AI-based image scene association optimization processing method takes into account that a large amount of data processing operations will be performed after the specific deployment of the solution. Therefore, the screening criterion is set in S3-4-1 as the number of restoration and verification processing is 3, which not only ensures that the cyclic iteration of the restoration and verification processing will not affect the overall implementation of the solution, but also avoids the error cycle caused by the error solution, and fails to obtain the target result.

[0123] In this embodiment, an image scene association optimization processing method based on AI is provided, in which the improvement measures in step S3-4 are to adjust the abnormal data of the current real-time image data according to the historical normal data value after screening and judgment, and analyze again with the help of the model after adjustment to achieve internal improvement of the solution and verify iterative processing.

[0124] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0126] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. An AI-based image scene association optimization processing method, characterized in that: include: S1, using real-time image data to obtain image scene correlation features; S1-1, collect real-time image data; S1-2, establishing static features and dynamic features of the real-time image data respectively according to the real-time image data; S1-2-1, collecting real-time image data to obtain image resolution and image target type corresponding to the real-time image data as static features of the real-time image data; S1-2-2, obtaining image exposure and image source signal according to the corresponding acquisition device of the real-time image data; S1-2-3, using the real-time image data to obtain image texture features; S1-2-4, using the image exposure, image source signal and image texture features as dynamic features of real-time image data; S1-3, using the static features and dynamic features of the real-time image data as image scene association features; S2. Using the real-time image data to establish an image scene association optimization model based on AI according to image scene association features; S2-1, using the real-time image data to establish an image target area recognition model based on AI according to the image scene association features corresponding to the dynamic features; S2-2, using the image scene association features to establish a basic image scene association optimization model based on AI; S2-3, using the image target area recognition model and the basic image scene association optimization model as the image scene association optimization model; S3. Obtaining an image scene association optimization processing result by using the image scene association optimization model.

2. The AI-based image scene association optimization processing method according to claim 1, characterized in that: Using the real-time image data to establish an image target area recognition model based on AI according to the image scene association features corresponding to the dynamic features includes: S2-1-1, using the real-time image data to establish a rough outline of an image target area of ​​the real-time image data according to image texture features corresponding to dynamic features of image scene association features; S2-1-2, obtaining a rough outline of an image target area corresponding to the historical image data according to an image source signal of a dynamic feature corresponding to an image scene association feature of the real-time image data; S2-1-3, using the rough outline of the image target area of ​​the historical image data to divide into a first training set; S2-1-4, using the rough outline of the image target area of ​​the real-time image data to divide into a first verification set; S2-1-5, using the first training set as input and the image area corresponding to the first training set as output, and establishing a basic image target area recognition model through AI-based training; S2-1-6, using the first verification set to input the basic image target area recognition model to perform model verification processing to obtain a basic image target area recognition verification output; S2-1-7, determine whether the basic image target area recognition verification output corresponds to the image texture feature of the real-time image data according to the image scene association feature corresponding to the dynamic feature, if so, execute S2-1-8. Otherwise, obtain the rough outline of the image target area of ​​the historical image data according to the image texture feature of the real-time image data according to the image scene association feature corresponding to the dynamic feature, and update the first training set, and return to S2-1-5; S2-1-8. Determine whether the image exposure corresponding to the basic image target area recognition verification output is consistent with the image exposure of the dynamic features corresponding to the image scene associated features of the real-time image data. If so, use the basic image target area recognition model as the image target area recognition model. Otherwise, obtain the rough outline of the image target area of ​​the historical image data according to the image exposure of the dynamic features corresponding to the image scene associated features of the real-time image data, update the first training set, and return to S2-1-5.

3. The AI-based image scene association optimization processing method according to claim 2, characterized in that: Using the image scene association features to establish a basic image scene association optimization model based on AI includes: S2-2-1, using the dynamic characteristics of the image scene association characteristics to correspond to the signal frequency of the image source signal to respectively obtain historical normal image data and historical abnormal image data; S2-2-2, obtaining abnormal pixels corresponding to the historical abnormal image data; S2-2-3, obtaining the image exposure, image source signal and image texture features corresponding to the dynamic features of the historical abnormal image data to respectively establish a first data set, a second data set and a third data set; S2-2-4. Using the first data set as input and the abnormal pixels corresponding to the first data set as output, an abnormal exposure recognition model is established through training based on AI; S2-2-5. Using the second data set as input and the abnormal pixels corresponding to the second data set as output, an abnormal source signal recognition model is established through training based on AI; S2-2-6. Using the third data set as input and the abnormal pixels corresponding to the third data set as output, an abnormal texture feature recognition model is established through training based on AI; S2-2-7. Use the abnormal exposure recognition model, abnormal source signal recognition model and abnormal texture feature recognition model to perform model output verification processing based on historical normal image data to obtain a basic image scene association optimization model.

4. The AI-based image scene association optimization processing method according to claim 3, characterized in that: The abnormal exposure recognition model, the abnormal source signal recognition model and the abnormal texture feature recognition model are used to perform model output verification processing according to historical normal image data to obtain a basic image scene association optimization model, including: S2-2-7-1, according to the dynamic characteristics of the image scene associated characteristics corresponding to the historical normal image data, respectively obtain the image exposure, image source signal and image texture characteristics to establish an exposure verification set, a source signal verification set and a texture feature verification set in turn; S2-2-7-2, using the exposure verification set to input the abnormal exposure recognition model to obtain the model output of the abnormal exposure recognition model; S2-2-7-3, using the source signal verification set to input the abnormal source signal recognition model to obtain the model output of the abnormal source signal recognition model; S2-2-7-4, using the texture feature verification set to input the abnormal texture feature recognition model to obtain the model output of the abnormal texture feature recognition model; S2-2-7-5, determine whether the model output of the abnormal exposure recognition model exists, if so, use the exposure verification set to update the first data set and return to S2-2-4, otherwise, execute S2-2-7-6; S2-2-7-6, determine whether the model output of the abnormal source signal identification model exists, if so, use the source signal verification set to update the second data set and return to S2-2-5, otherwise, execute S2-2-7-7; S2-2-7-7. Determine whether the model output of the abnormal texture feature recognition model exists. If so, use the texture feature verification set to update the third data set and return to S2-2-6. Otherwise, use the abnormal exposure recognition model, abnormal source signal recognition model and abnormal texture feature recognition model as the basic image scene association optimization model.

5. The AI-based image scene association optimization processing method according to claim 3, characterized in that: The image scene association optimization processing result obtained by using the image scene association optimization model includes: S3-1, using the image target area recognition model of the real-time image data input image scene association optimization model to respectively obtain the image target area of ​​the real-time image data and the image non-target area of ​​the real-time image data; S3-2, using the dynamic features of the image scene association features corresponding to the real-time image data to input the basic image scene association optimization model of the image scene association optimization model to obtain abnormal exposure recognition results, abnormal source signal recognition results and abnormal texture feature recognition results respectively; S3-3, using the image target area and the image non-target area of ​​the real-time image data to perform restoration verification processing according to the static features of the image scene associated features to obtain a restoration verification result of the real-time image data; S3-4. Utilizing the restoration verification result of the real-time image data, the image scene association optimization processing result is obtained according to the abnormal exposure recognition result, the abnormal source signal recognition result and the abnormal texture feature recognition result.

6. The AI-based image scene association optimization processing method according to claim 5, characterized in that: The restoration verification result of the real-time image data obtained by performing restoration verification processing on the image target area and the image non-target area of ​​the real-time image data according to the static features of the image scene associated features includes: S3-3-1, determine whether the image target area, image non-target area and static features of the image scene associated features of the real-time image data correspond to the same image target type, if so, execute S3-3-2, otherwise, return to S1-2-1; S3-3-2. Determine whether the image resolution corresponding to the image target area of ​​the real-time image data, the image resolution corresponding to the image non-target area, and the image resolution corresponding to the static features of the image scene associated features are all consistent. If so, the restoration verification result of the real-time image data is passed. Otherwise, the restoration verification result of the real-time image data is failed, and the real-time image data is updated and returned to S1-1.

7. The AI-based image scene association optimization processing method according to claim 6, characterized in that: The image scene association optimization processing results are obtained by using the restoration verification results of the real-time image data according to the abnormal exposure recognition results, the abnormal source signal recognition results and the abnormal texture feature recognition results, including: S3-4-1, determine whether the number of restoration verification processing corresponding to the restoration verification result of the real-time image data is greater than 3, if so, return to S2-1-1, otherwise, execute S3-4-2; S3-4-2, determine whether the abnormal exposure recognition result exists, if so, use the image exposure of the dynamic feature corresponding to the historical normal image data to adjust the real-time image data according to the abnormal exposure recognition result, and return to S3-2, otherwise, execute S3-4-3; S3-4-3, determine whether the abnormal source signal identification result exists, if so, use the image source signal of the dynamic feature corresponding to the historical normal image data to perform noise reduction processing on the image source signal of the dynamic feature corresponding to the real-time image data, and return to S3-2, otherwise, execute S3-4-4; S3-4-4, determine whether the abnormal texture feature recognition result exists, if so, use the image texture features of the dynamic features corresponding to the historical normal image data to update the real-time image data according to the abnormal texture feature recognition result, and return to S3-2, otherwise, execute S3-4-5; S3-4-5. Determine whether the real-time image data at the current moment and the image target area and image non-target area corresponding to the restoration verification result are consistent. If so, use the real-time image data at the current moment as the image scene association optimization processing result. Otherwise, use the real-time image data at the current moment for update processing and return to S1-2.

Images