Agent-based image analysis and processing system and method, and storage medium

Abstract

The present invention relates to the technical field of computer image processing, and in particular to an agent-based image analysis and processing method and system, and a storage medium. Image quality analysis and estimation is performed by means of a pre-trained image quality analysis model, and a corresponding processing strategy and a corresponding image processing sequence are generated, thereby achieving automatic image quality analysis and evaluation and facilitating selection of different processing strategies based on different image content, so as to satisfy processing requirements of various application scenarios. A corresponding image processing model is retrieved on the basis of the result of matching between the processing strategy and basic information of each image processing model in a knowledge base, and then called according to the image processing sequence, and an automatic image processing flow control model automatically controls the image processing model to execute image processing according to the image processing sequence and performs monitoring, thereby achieving automatic image processing, reducing the requirement of manual intervention, and saving the time and cost.

Description

An agent-based image analysis and processing system, method and storage medium TECHNICAL FIELD

[0001] The present application belongs to the technical field of image analysis and processing, and specifically relates to an agent-based image analysis and processing system, method and storage medium. BACKGROUND

[0002] In the digital era, image processing and optical character recognition (OCR) technologies play a crucial role in document management, automated office work, intelligent monitoring, and other fields. Traditional image processing methods rely on manual intervention, such as manually adjusting image brightness, contrast, and noise reduction, to improve the accuracy of OCR. However, this method is inefficient and difficult to maintain consistent processing quality when faced with a large number of images. In addition, with the popularization of multilingual environments, higher requirements are placed on image processing systems, which need to be able to recognize and process text in different languages, further increasing the complexity of processing.

[0003] Existing automatic image processing systems have reduced the burden of manual intervention to some extent, but they often lack intelligent judgment mechanisms and cannot automatically select the most appropriate processing strategy based on the specific content and quality of the image. These systems may apply the same processing flow to all images, resulting in poor processing results in some cases and even damaging image quality, which cannot meet the processing needs of various application scenarios.

[0004] SUMMARY

[0005] To solve the technical problems existing in the background art, the present application provides an agent-based image analysis and processing system and method, aiming to realize automatic analysis of image quality, intelligent selection of appropriate image processing strategies, and execution of automated image processing procedures, thereby helping to improve image processing quality and meet the processing needs of various application scenarios, while also helping to reduce the need for manual intervention, saving time and cost. The specific steps of the present application are as follows:

[0006] To achieve the above technical solution, in a first aspect, a 1. agent-based image analysis and processing method is provided, comprising:

[0007] Step one, based on a general multi-model base large model, an image quality analysis model is constructed, and the image quality analysis model is trained;

[0008] Step two, input the image to be processed;

[0009] Step three, evaluate the quality of the input image, and based on the evaluation results, determine the processing strategy and the image processing sequence corresponding to the processing strategy;

[0010] Step three, based on the processing strategy and the basic information of each image processing model recorded in the knowledge base, retrieve the corresponding image processing model from the image processing model library;

[0011] Step four, call the retrieved image processing model according to the image processing order;

[0012] Step five, automatically control the image processing model to implement the image processing process according to the image processing order;

[0013] Step six, output the processed image.

[0014] Further, the method further comprises:

[0015] According to the demand, it is judged whether the language and characters in the image need to be recognized, if yes, the processed image is automatically subjected to language and character recognition, and the recognized text is output.

[0016] Further, the quality assessment of the input image comprises: analyzing the basic properties of the image, identifying the detailed features of the image, recognizing the language and characters of the image, and identifying the detailed features of the language and characters in the image based on the recognized language and characters.

[0017] Further, the step five further comprises:

[0018] Automatically monitor the image processing process and record the problems in the processing process;

[0019] According to the image processing process, allocate management resources.

[0020] Further, the method further comprises: viewing the processing progress, viewing the recorded problems and comparing the images before and after processing, based on the recorded problems and the comparison results, inputting feedback;

[0021] Periodically learn feedback to update the image processing model library and the knowledge base.

[0022] On the other hand, the present application provides an image analysis and processing system based on an intelligent agent, comprising:

[0023] An image quality analysis model construction and training module is used to construct an image quality analysis model based on a general multi-model base large model, and to train the image quality analysis model;

[0024] An input module is used to input an image;

[0025] An image quality analysis model is used to assess the quality of the input image, and based on the assessment result, determine a processing strategy and an image processing order corresponding to the processing strategy;

[0026] The plurality of image processing models are used to process a plurality of image problems.

[0027] The image processing model library is used to store the plurality of image processing models.

[0028] The knowledge base is used to record basic information of each model in the plurality of image processing models, and the basic information includes applicable scenarios, input and output formats, and effect descriptions.

[0029] The retrieval model is used to retrieve corresponding image processing models from the image processing model library based on the processing strategy and the basic information of each model.

[0030] The calling model is used to call the retrieved image processing models according to the image processing sequence.

[0031] The automated image processing flow control model is used to automatically control the image processing models to implement the processing of the image according to the image processing sequence.

[0032] The output model is used to output the processed image.

[0033] Further, the automated image processing flow control model further includes a monitoring unit and a resource allocation unit.

[0034] The monitoring unit is used to monitor the processing of the image to record problems in the processing.

[0035] The resource allocation unit is used to manage resource allocation.

[0036] Further, the input module includes a user interaction and feedback module and an application programming interface.

[0037] The user interaction and feedback module is used to upload images, view processing progress, view recorded problems, compare images before and after processing, and input feedback.

[0038] The application programming interface is used to batch upload images.

[0039] Further, the system further includes a continuous learning and optimization module for optimizing the image processing model and the knowledge base based on feedback.

[0040] Further, the plurality of image processing models include:

[0041] The shadow removal model is used to remove shadows in the image.

[0042] The clarity enhancement model is used to enhance the clarity of the image.

[0043] The binarization processing model is used to perform binarization processing on the image.

[0044] a text sharpening model configured to sharpen text in the image;

[0045] an OCR model configured to recognize text in the image.

[0046] In yet another aspect, the present application provides a computer-readable storage medium comprising a stored program, wherein the program, when executed, controls a device in which the computer-readable storage medium is located to perform the above-mentioned image analysis and processing method based on an agent.

[0047] The present application has the following beneficial effects:

[0048] (1) The present application analyzes and evaluates the image quality through the pre-trained image quality analysis model, and generates the corresponding processing strategy and the processing flow corresponding to the processing strategy based on the quality evaluation result, thereby realizing the automatic analysis and evaluation of the image quality and helping to realize the selection of different processing strategies based on different image contents to meet the processing needs of various application scenarios.

[0049] (2) According to the matching result of the processing strategy and the basic information of each image processing model in the knowledge base, the corresponding image processing model is retrieved from the image processing model library according to the image processing sequence, and the image processing model is called according to the image processing sequence, the automatic image processing flow control model automatically controls the image processing model to perform the image processing process according to the image processing sequence, and the image processing process is monitored, thereby helping to realize the automatic processing of the image and reducing the need for manual intervention, saving time and cost.

[0050] (3) The image quality analysis model of the present application can also recognize multiple languages and characters, which helps to significantly improve the accuracy of optical character recognition (OCR).

[0051] (4) The present application also learns from feedback regularly to improve the accuracy and efficiency of future image processing.

[0052] (5) The present application can also upload images in batches through an application programming interface (API), improving the processing efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0053] The drawings accompanying the specification of the present application serve to provide a further understanding of the present application, and the illustrative embodiments of the present application and their descriptions serve to explain the present application, and do not constitute an improper limitation on the present application.

[0054] Fig. 1 is a flowchart of one embodiment of the image analysis and processing method based on an agent of the present application. DETAILED DESCRIPTION

[0055] The application will be further described below with reference to the drawings and embodiments.

[0056] It should be noted that the following detailed description is illustrative only and is intended to provide further description of the application. Unless otherwise defined, each technical and scientific term used herein has the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0057] It should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and it should be understood that when the terms "comprise" and / or "include" are used in the specification, they refer to the presence of a feature, step, operation, device, component, and / or combination thereof.

[0058] In the present application, the terms such as "upper", "lower", "left", "right", "front", "back", "vertical", "horizontal", "side", "bottom", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, which is only a relationship word determined for the purpose of describing the structural relationship of the components or elements of the present application, and cannot be understood as a limitation on the present application.

[0059] In the present application, the terms such as "fixedly connected", "connected", "connected" should be understood broadly, which means that it can be fixedly connected, integrally connected or detachably connected; it can be directly connected or indirectly connected through an intermediate medium. For relevant researchers or technicians in the art, the specific meaning of the above terms in the present application can be determined according to the specific circumstances, and cannot be understood as a limitation on the present application.

[0060] Example 1

[0061] As shown in FIG. 1, the present embodiment provides an image analysis and processing method based on an agent, which includes the following steps:

[0062] S1: Construct and train an image quality analysis model.

[0063] A general multi-model base large model is selected, and in this case, an open source cogVLM model is selected. The lora or lora fast fine-tuning method is used to fine-tune the model, create an image quality analysis model, and input the image to be analyzed and the corresponding prompt. The output is the quality problem of the image and the corresponding language. The model weight can be trained to obtain better performance in the task. The training and application of the image quality analysis large model and the prompt are as follows:

[0064] # Role

[0065] A distinguished OCR solution expert with deep expertise can easily help users design and implement an OCR file recognition workflow.

[0066] ##Objective

[0067] Design a complete process for users to implement OCR file recognition based on images. According to the uploaded OCR interface document, clearly indicate the order and method of interface calls. No specific OCR file processing is required, but the order of interface calls and specific URL and method calls must be specified. Only use the URLs in the uploaded document. Note the difference between using table recognition and OCR recognition, and determine whether to perform preprocessing. Finally, return the processed workflow and the corresponding URL in the document.

[0068] ##Skills

[0069] ###Skill 1: Accurate understanding of OCR interface

[0070] - Analyze and accurately understand the OCR interface document to find the order of interface calls, including preprocessing interfaces and OCR interfaces. Can clearly indicate the URL and calling method of the interface.

[0071] ###Skill 2: Determine image preprocessing needs

[0072] - Determine whether image preprocessing is needed before OCR processing based on the needs of the OCR solution, and determine whether image correction and other work are needed, and call the relevant preprocessing interfaces.

[0073] ###Skill 3: Return the exact URL in the uploaded document

[0074] - According to the process, ensure that the exact URL in the uploaded document is returned, and ensure that no URL in the document is returned.

[0075] S2: Input image.

[0076] Upload the image to be processed through user interaction and feedback module, or upload the image in bulk through application programming interface (API), and the system will convert the received image to be processed into a unified format, such as png format, to facilitate subsequent processing.

[0077] S3: Quality analysis of image.

[0078] Specifically, the image quality analysis model analyzes basic attributes of the image such as sharpness and color balance, recognizes the language and characters in the image, and identifies the shadow in the image and based on the recognized language and characters, recognizes the depth and legibility of the language and characters in the image and other detailed features, and evaluates the analyzed basic attributes and recognized detailed features to generate an evaluation result, and then determines a processing strategy and an image processing sequence corresponding to the processing strategy based on the evaluation result. The evaluation result includes quality problems of the image; the quality problems include key features of the image, and the key features can be presented in the form of a list, and the main languages include but are not limited to Chinese, English, and Tibetan, etc.

[0079] S4: retrieve the corresponding image processing model.

[0080] The processing strategy is matched with the basic information of each image processing model recorded in the knowledge base, so as to retrieve the required image processing model from the image processing model library according to the matching result.

[0081] Specifically, the processing strategy includes scene, format, and expected effect. For example, if the image has a shadow and the text legibility is low, the generated image processing strategy may be shadow removal and text sharpening, then the matching result is a shadow removal model with an effect description of shadow removal and a text sharpening model with an effect description of text sharpening. Then the shadow removal model and the text sharpening model are retrieved from the image processing model library according to the matched shadow removal model and the text sharpening model.

[0082] S5: automatically execute image processing.

[0083] According to the image processing sequence, the retrieved models are called, and the images are processed in sequence according to the calling sequence, that is, each image processing model receives the image and processing parameters processed by the previous image processing model, performs a specific image enhancement task, and after processing is completed, the processed image is passed to the next processing model until all the called image processing models are processed.

[0084] For example, if the image has a shadow and the text legibility is low, the shadow removal model is called first, and then the text sharpening model is called, and the image processing process is that after the image is processed by the shadow removal model, the image after shadow removal is passed to the text sharpening model, and the image is processed by the text sharpening model to complete the processing of the image.

[0085] S6: automatically monitor the image processing flow.

[0086] The automatic image processing flow control model monitors the entire processing process to ensure that each image processing model is executed in sequence, and any abnormality occurring in the processing process is recorded and processed.

[0087] The automated image processing flow control model is also responsible for managing resource allocation, ensuring that the system can efficiently process a large number of images while avoiding overload.

[0088] At the same time, the automated image processing flow control model will also choose whether to call the OCR interface according to user needs, automatically perform text recognition from the processed image according to the recognized language and text, and return the recognized text.

[0089] S7: Output the processed image and feedback the processing effect.

[0090] After all the called image processing models complete processing, the system will output the final processed and optimized image to the user-specified location, such as local storage or cloud storage.

[0091] Users can view the images before and after processing through the user interaction and feedback module, and compare them to evaluate the processing effect and provide feedback.

[0092] At the same time, users can also view the problems recorded during the image processing process through the user interaction and feedback module to provide feedback based on the recorded problems.

[0093] S8: Continuous learning and optimization.

[0094] The continuous learning and optimization module regularly learns user feedback, updates the processing strategies and effect descriptions in the knowledge base, to improve the accuracy and efficiency of future image processing.

[0095] Embodiment 2:

[0096] The embodiment provides an image analysis and processing system based on an agent, which comprises:

[0097] (I) Image quality analysis model construction and training module: used to build an image quality analysis model capable of comprehensively evaluating the quality of input images based on a general multi-model base large model (such as cogVLM model in this example), and train the image quality analysis model.

[0098] Specifically, fine-tuning is performed on the basis of the general multi-model base large model, and the fine-tuning method adopts lora or Q-lora fast fine-tuning method, the input is the image to be analyzed and the standard prompt, and the output is the quality problem of the image, the corresponding processing strategy and the processing order corresponding to the processing strategy. Train the model weight, so that better performance can be obtained in the task.

[0099] The general multi-model base large model includes an MLP layer and a transformer layer, the general multi-model base large model is fine-tuned by using the lora or Q-lora fast fine-tuning method to obtain the image quality analysis model, and in fact, the parameters outside the transformer layer are frozen, that is, different characteristic images and standard prompts are input, the model output is compared with the correct result, that is, the text describing the image characteristics, the corresponding processing strategy and the processing order corresponding to the processing strategy, the loss is calculated and the parameters of the transfermer layer are updated.

[0100] (ii) Input module: The input module includes a user interaction and feedback module and an application programming interface.

[0101] The user interaction and feedback module is not only used for uploading images, but also for viewing the processing progress, comparing the effects before and after processing, and the user can input feedback according to the comparison results.

[0102] The application programming interface (API) is used for batch uploading of images, thereby improving the efficiency of uploading images.

[0103] In addition, a conversion unit for converting images into a unified format is provided in the input module.

[0104] (iii) The image quality analysis model is used to analyze the basic properties of the image such as sharpness and color balance, recognize the language and characters of the image, and identify the shadows in the image and the details such as the depth and recognizability of the language and characters in the image based on the recognized language and characters, and evaluate the basic properties and details identified to generate an evaluation result, and then determine the processing strategy and the image processing order corresponding to the processing strategy based on the evaluation result. The evaluation result includes the quality problems of the image; the quality problems include the key features of the image, and these key features can be presented in the form of a list, and the main languages include but are not limited to Chinese, English and Tibetan, etc.

[0105] Specifically, as described in (i), the image quality analysis model selects cogVLM and is fine-tuned accordingly. Specifically, the image quality analysis model vectorizes the image and the text respectively, then aligns the image vector with the text vector through the forward propagation of the MLP layer, then fuses the two vectors of the text and the image, outputs to the transformer layer for processing, and finally outputs the image characteristics, wherein the image characteristics include the above-mentioned detail features and basic properties.

[0106] Specifically, the MLP layer is a double-layer forward propagation layer, and the MLP layer aligns the image vector with the text vector by using the SwiGLU activation function, and the specific formula is: h1=(W1*x+b1)⊙σ(Wg1 (W1*x+b1)+b g1 ) y=W2*h1+b2

[0107] where x is the input vector, h1 is the hidden layer vector, y is the output vector, represents element-wise multiplication, σ is the Sigmoid activation function, W1, b1 are the weight matrix and bias parameter of the first layer, W g1 , b g1 are the weight and bias of the SwiGLU activation function, W2, b2 are the weight matrix and bias parameter of the second layer.

[0108] (Four) multiple image processing models for realizing processing of multiple image problems.

[0109] The multiple image processing models include but are not limited to a shadow removal model for removing shadows in an image, a sharpness enhancement model for enhancing the sharpness of an image, a binarization processing model for binarization processing of an image, a text sharpening model for sharpening text in an image, and an OCR model for recognizing text in an image. Different training databases of different image processing models have different main languages, which gives different models different language biases and helps accurate processing of different languages. Moreover, each model can be encapsulated as an independent module with a standardized interface, facilitating system calling and integration.

[0110] The binarization processing model can use the confidence Sauvola binarization algorithm to add a confidence score to background pixels through a selective automatic encoder for document image binarization, to distinguish text and background, thereby realizing document binarization of light text and applying to the case of light document text. This embodiment takes the pixel of the image as P for example. First, the binarization threshold of the pixel p is calculated, and the specific formula is as follows:

[0111] where T W (p) represents the binarization threshold of the pixel p, which is calculated according to the window W around the pixel p. represents the average value of all pixels in the window W centered on the pixel p. represents the standard deviation of all pixels in the window W centered on the pixel p. R is a dynamic range defined as the difference between the maximum and minimum values of the input image divided by 2. k is a parameter in the range of [0, 1] for adjusting the calculation of the threshold.

[0112] Then, according to the calculated binarization threshold, the confidence is calculated as follows:

[0113] where I(p) represents the value of the pixel p in the input image I, TW (p) represents the binarization threshold of pixel p calculated using formula (1), where the window size is W, and max(I) represents the maximum pixel value of the input image I.

[0114] Specifically, if the value of pixel p is greater than the threshold calculated according to formula (1), the confidence is equal to (pixel value - threshold) / (image maximum value - threshold); otherwise, the confidence is 0.

[0115] That is, when the pixel value is greater than the threshold, it is considered a foreground pixel, and the confidence is a normalized value. The numerator is the difference between the pixel value and the threshold, and the denominator is the difference between the image maximum value and the threshold. In this way, the confidence is limited to between 0 and 1. Otherwise, it is considered a background pixel, and the confidence is directly set to 0.

[0116] After calculating the confidence of each pixel of the input image using the above formula, a new confidence image can be obtained, and the confidence image can be binarized to obtain the final binarization result image.

[0117] The shadow removal model can use a shadow removal machine learning model that combines background estimation and U-Net to remove shadows from documents with obvious shadows, thereby solving the problem of correct text being disturbed by obvious shadows.

[0118] Specifically, the shadow removal model of this example uses a BEDSR-Net model structure, which is divided into two modules. The first module uses neural network feature extraction to obtain the background color of the image, and the second module uses a conditional generative adversarial network to identify and remove shadows based on the background color.

[0119] In addition, the various image processing models can also include a machine learning diffusion model for denoising documents, removing noise points in documents, and a machine learning model for simultaneously modeling two-dimensional and three-dimensional documents, for correcting and flattening documents, thereby eliminating the effects of document distortion.

[0120] It should be noted that these image processing models can be increased or decreased according to actual needs to adapt to different application scenarios.

[0121] (Five) Image processing model library, used to store a variety of image processing models.

[0122] (Six) Knowledge base, used to record the basic information of each model in the various image processing models in detail, including the applicable scenarios, input and output formats, and effect descriptions. The effect description is the function of the model, for example, the effect description of the clarity enhancement model is to enhance the clarity of the image, and the effect description of the shadow removal model is to remove the shadows in the image, and so on.

[0123] The knowledge base is a built-in external knowledge base of the system, and the embodiment uses a Milvus database.

[0124] (Seven) Retrieval model, for retrieving corresponding image processing models from the image processing model library based on the processing strategy and the basic information of each image processing model.

[0125] Specifically, the evaluation results of the image quality analysis model are matched with the recorded basic information of each image processing model in the knowledge base, so as to retrieve the required image processing models from the image processing model library according to the matching results.

[0126] (Eight) Call model, for calling the retrieved image processing models in the image processing order.

[0127] Specifically, the retrieved models are called in the image processing order. For example, if the image has shadows and the text is not easily recognizable, the de-shadowing model is called first, and then the text sharpening model is called.

[0128] (Nine) Automatic image processing flow control model, for automatically controlling the image processing models to implement the processing process of the image in the image processing order, realizing the full automation of image processing.

[0129] Specifically, the automatic image processing flow control model not only manages the flow of images between various image processing models, but also monitors the entire processing process to ensure that each image processing model is executed in order, and any abnormality in the processing process can be discovered and processed in time.

[0130] In addition, the automatic image processing flow control model also supports batch processing, greatly improving the efficiency of processing a large number of images. At the same time, the automatic image processing flow control model is also responsible for managing resource allocation to ensure that the system can efficiently process a large number of images and avoid overload.

[0131] Finally, the automatic image processing flow control model will also select whether to call the OCR interface according to user needs, automatically perform text recognition on the processed image, and return the recognized text.

[0132] (Ten) Output model, for outputting the processed image.

[0133] Specifically, after all image processing steps are completed, the output model outputs the most processed image to the user-specified location, such as local storage or cloud storage.

[0134] (Eleven) Continuous learning and optimization module, for optimizing the image processing models and the knowledge base based on feedback.

[0135] Specifically, the continuous learning and optimization module learns user feedback periodically, updates the processing strategy and effect description in the knowledge base, so as to improve the accuracy and efficiency of future image processing.

[0136] It should be noted that the method of embodiment 1 can be implemented by relying on the system provided in embodiment 2.

[0137] Embodiment 3:

[0138] The embodiment provides a computer-readable storage medium, the computer-readable storage medium comprises a stored program, wherein the program controls the device where the computer-readable storage medium is located to execute the agent-based image analysis and processing method in embodiment 2 when the program is running.

[0139] In the specification, the same or similar parts between various embodiments can be referred to each other. Especially, for the terminal embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the related parts can be referred to the description in the method embodiment.

[0140] In several embodiments provided in the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic, for example, the division of the unit is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be through some interface, indirect coupling or communication connection between the system or unit, and can be electrical, mechanical or other forms.

[0141] The unit described as a separate component can or can not be physically separated, and the component displayed as a unit can or can not be a physical unit, that is, it can be located in one place, or can be distributed on a plurality of network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the embodiment scheme.

[0142] In addition, it should be noted that the flowcharts in the drawings show the methodology of the embodiments of the present disclosure, and the operations or steps corresponding to the different blocks in the flowcharts or block diagrams in the drawings can also occur in an order different from that disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, which can depend on the functions involved. Each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0143] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An agent-based image analysis and processing method, characterized by, The method comprises the following steps: Step 1: constructing an image quality analysis model based on a general multi-model base large model, and training the image quality analysis model; Step 2: inputting an image to be processed; Step 3: evaluating the quality of the input image, and determining a processing strategy and an image processing sequence corresponding to the processing strategy based on the evaluation result; Step 4: retrieving corresponding image processing models from an image processing model library based on the processing strategy and basic information of each image processing model recorded in a knowledge base; Step 5: calling the retrieved image processing models according to the image processing sequence; Step 6: automatically controlling the image processing models to implement the processing procedure of the image according to the image processing sequence; Step 7: outputting the processed image.

2. The agent-based image analysis and processing method of claim 1, wherein, The method further comprises the following steps: According to the requirement, it is judged whether the language and characters in the image need to be recognized, if yes, the processed image is automatically subjected to language and character recognition, and the recognized text is outputted.

3. The agent-based image analysis and processing method of claim 1, wherein, The quality evaluation of the input image comprises the following steps: analyzing the basic attributes of the image, recognizing the detailed features of the image, recognizing the language and characters of the image, and recognizing the detailed features of the language and characters in the image based on the recognized language and characters.

4. The agent-based image analysis and processing method of claim 1, wherein, The step 5 further comprises the following steps: Automatically monitoring the processing procedure of the image, and recording problems in the processing procedure; According to the processing procedure of the image, resources are allocated and managed.

5. The agent-based image analysis and processing method of claim 4, wherein, The method further comprises the following steps: checking the processing progress, checking the recorded problems, and comparing the image before and after processing, inputting feedback based on the recorded problems and the comparison result; Periodically learning the feedback to update the image processing model library and the knowledge base.

6. An agent-based image analysis and processing system, characterized by, The method comprises the following steps: An image quality analysis model construction and training module is used to construct an image quality analysis model based on a general multi-model base large model, and train the image quality analysis model; An input module is used to input an image; An image quality analysis model is used to evaluate the quality of the input image, and determine a processing strategy and an image processing sequence corresponding to the processing strategy based on the evaluation result; A plurality of image processing models are used to implement the processing of a plurality of image problems; An image processing model library is used to store the plurality of image processing models; A knowledge base is used to record the basic information of each model in the plurality of image processing models, and the basic information comprises applicable scenarios, input and output formats, and effect descriptions; A retrieval model is used to retrieve corresponding image processing models from the image processing model library based on the processing strategy and the basic information of each model; A calling model is used to call the retrieved image processing models according to the image processing sequence; An automatic image processing flow control model is used to automatically control the image processing models to implement the processing procedure of the image according to the image processing sequence; An output model is used to output the processed image.

7. The agent-based image analysis and processing system of claim 6, wherein, The automatic image processing flow control model further comprises a monitoring unit and a resource allocation unit; The monitoring unit is used to monitor the processing procedure of the image to record problems in the processing procedure; The resource allocation unit is used to manage resource allocation.

8. The agent-based image analysis and processing system of claim 6, wherein, The input module comprises a user interaction and feedback module and an application programming interface. The user interaction and feedback module is configured to upload images, view processing progress, view recorded problems, compare images before and after processing, and input feedback. The application programming interface is configured to batch upload images. 9.The agent-based image analysis and processing system according to claim 6, wherein, The system further comprises a continuous learning and optimization module configured to optimize the image processing model and the knowledge base based on feedback.

10. The agent-based image analysis and processing system of claim 6, wherein, The plurality of image processing models comprises: A shadow removal model configured to remove shadows in the images; An image sharpness enhancement model configured to enhance sharpness of the images; A binarization processing model configured to perform binarization processing on the images; A text sharpening model configured to sharpen text in the images; An OCR model configured to recognize text in the images.

Images