AI-based microscopic image creative generation system and method

Abstract

This invention provides an AI-based creative generation system and method for microscopic images, relating to the fields of artificial intelligence and image processing technology. The system comprises an image input module, a feature analysis module, an AI generation module, a human-computer interaction module, and an output module. First, the image input module acquires real microscopic images. The feature analysis module extracts and parameterizes the structured scientific features of the images. Then, the human-computer interaction module obtains simple sketches or text prompts from children, which, along with the parameterized features, serve as control conditions and are input into a generative AI model fine-tuned from the microscopic image dataset. Through the fusion of structured extraction of scientific features and constraints generated by AI, child-friendly interdisciplinary interactive design, and deep coupling of science education and aesthetic education, the system both encourages children to observe the details of scientific specimens in depth and inspires a sense of accomplishment through personalized creative output, achieving a seamless connection between scientific observation and artistic expression. This has significant educational value and application prospects.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and image processing technology, specifically to an AI-based microscopic image creative generation system and method. Background Technology

[0002] One of the core goals of children's science education is to cultivate observation skills and innovative thinking. However, scientific observation and artistic creation are currently mostly independent activities, lacking effective interdisciplinary transformation tools.

[0003] Scientific observation is superficial: When children observe under a microscope, they focus more on "morphological recognition" and lack the ability to explore deeper scientific features such as texture and symmetry, resulting in insufficient depth of observation. Artistic creation lacks scientific support: Children's art education relies heavily on abstract imagination or fixed templates, making it difficult to transform scientific observations into creative materials, and the cultivation of interdisciplinary thinking is lacking. Insufficient tool adaptability: Existing AI image generation tools mostly rely on plain text or sketch input, without combining scientific observation scenarios, and cannot transform the scientific characteristics of real microscopic images into constraints for creative expression; Limited interactive experience: It lacks interactive design tailored to children's cognitive level, fails to balance the accuracy of scientific features with the flexibility of creative expression, and is difficult to stimulate children's sustained enthusiasm for participation.

[0004] Existing technologies have failed to address the coordination issues of "scientific feature extraction - creative expression integration - child-friendly interaction," leading to a disconnect between science education and art education and failing to fully tap into children's interdisciplinary potential.

[0005] Therefore, developing an AI generation system that can deeply integrate the scientific features of real microscopic images with children's creativity has become an urgent need to promote the coordinated development of children's science education and aesthetic education. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an AI-based microscopic image creative generation system and method, aiming to solve the problems of separation between children's scientific observation and artistic creation, and the lack of interdisciplinary transformation tools.

[0007] To achieve the above objectives, the present invention provides the following technical solution: An AI-based microscopic image creative generation system, comprising an image input module, a feature analysis module, an AI generation module, a human-computer interaction module, and an output module; The image input module includes a microscope interface unit, an image uploading unit, and a preprocessing unit; The microscope interface unit supports connection to a digital microscope via USB and wireless interface to directly acquire real-time microscopic images; The image uploading unit supports uploading local microscopic image files; The preprocessing unit uses 3×3~5×5 Gaussian filtering to remove image noise, adjusts brightness and contrast through histogram equalization, and normalizes the image size to 512×512 pixels to provide standardized input for feature extraction. The feature analysis module includes a texture extraction unit, a symmetry analysis unit, a shape contour extraction unit, a color distribution statistics unit, and a parameterization unit; The texture extraction unit uses the gray-level co-occurrence matrix (GLCM) to calculate four parameters of the image in the 0°, 45°, 90°, and 135° directions, namely energy, entropy, contrast, and correlation, to comprehensively characterize the texture features of the microscopic image. The symmetry analysis unit transforms the image to the frequency domain through Fourier transform, calculates the similarity scores of left-right, top-bottom, and central symmetry, and quantifies the symmetry features. The shape contour extraction unit uses the Canny edge detection algorithm to extract the target contour, and combines the Hu moment to calculate 7 invariant moment parameters to characterize the shape's morphological features, while outputting the contour coordinate point set. The color distribution statistics unit converts the image from RGB space to HSV space, statistically analyzes the probability distribution histograms of hue (H), saturation (S), and brightness (V), and extracts peak, mean, and variance parameters. The parameterization unit normalizes the parameters of the above four types of features to the range of 0 to 1, and splices them together to form a 128-dimensional feature control set, which supports dynamic adjustment of feature weights. The AI ​​generation module includes a model fine-tuning unit and a constraint generation unit; The model fine-tuning unit is based on the Stable Diffusion V1.5 or V2.1 framework and uses LoRA lightweight fine-tuning technology. It is trained using a dataset of 10,000+ microscopic images covering biological (e.g., cells, microorganisms), chemical (e.g., crystals, reagent reactions), and physical (microstructure) fields. The backbone network parameters are frozen, and only the low-rank matrix is ​​trained to ensure that the model's reproduction accuracy of scientific features is ≥90%. The constraint generation unit fuses the feature control set with creative control signals (such as sketch outline vectors or text keyword vectors), inputs them into the fine-tuned model, and achieves deep coupling between scientific features and creative expression through a cross-attention mechanism to generate creative images. The generation time is ≤10s, and the resolution is adjustable from 512×512 to 2048×2048. The human-computer interaction module includes a sketching unit, a text input unit, and an interaction control unit; The sketching unit provides a simple drawing interface, supports pressure-sensitive pen strokes and basic graphic tools (such as lines, circles, and triangles). After children draw, the sketch outline is extracted through threshold segmentation and morphological processing to generate outline control vectors. The text input unit supports Pinyin and simplified and traditional Chinese input, has a built-in vocabulary database for children, and uses a lightweight BERT model to parse text prompts into 256-dimensional keyword vectors. The interactive control unit supports feature weight adjustment, image resolution selection, and secondary optimization triggering, adapting to children's operating habits. The output module includes an image rendering unit, an annotation generation unit, and a storage and sharing unit; The image rendering unit performs color optimization and edge smoothing on AI-generated images to improve visual effects. The annotation generation unit can automatically generate a "scientific feature annotation map", which uses different colors to annotate the scientific features such as texture, shape and color in the creative image, and includes a brief text description; The storage and sharing unit supports saving creative images and labeled images locally in JPG, PNG, and PDF formats, and supports sharing to mobile phones and printers via Bluetooth and Wi-Fi for rapid output.

[0008] Furthermore, the image input module supports direct acquisition of microscopic images and local image upload. Preprocessing uses Gaussian filtering for noise reduction, and normalization mapping maps image pixel values ​​to the 0~1 range to ensure the consistency of image brightness and contrast.

[0009] Furthermore, the output module supports local saving, printing, and sharing of creative images, while generating a "scientific feature annotation map" to annotate the scientific features of the microscopic images integrated into the creative images, enabling scientific tracing of the creative process.

[0010] Furthermore, this AI-based method for generating creative microscopic images includes the following steps: Step 1: Microscopic Image Acquisition and Preprocessing The image input module acquires real microscopic images, and the images are preprocessed by noise reduction and normalization to ensure the accuracy of scientific feature extraction. Step 2: Structural Extraction and Parameterization of Scientific Features The feature analysis module performs structured analysis on the preprocessed microscopic images, extracting four core scientific features: texture, symmetry, shape contour, and color distribution. Each feature is then converted into a quantifiable parameter vector to form a feature control set. Step 3: Acquiring Creative Input from Children The human-computer interaction module receives simple sketches or text prompts from children, extracts the outlines of the sketches, performs semantic analysis on the text prompts, and generates creative control signals. Step 4: AI Model Constraint Generation The feature control set and creative control signal are input into a generative AI model finely tuned by a microscopic image dataset. The model integrates creative expression under scientific feature constraints to generate creative images. Step 5: Creative Image Output and Optimization The output module outputs creative images, and children can adjust the creative input or feature weights through the human-computer interaction module to achieve personalized optimization.

[0011] Furthermore, the algorithm for extracting the scientific features mentioned in step 2 is as follows: Texture features are extracted using the Gray-Level Co-occurrence Matrix (GLCM) to obtain energy, entropy, and contrast parameters. Symmetry features are analyzed using Fourier transform to determine the similarity between left-right and top-bottom symmetries of images. The shape contour features are extracted using Hu moments and the Canny algorithm for edge detection to extract contour coordinates and morphological parameters. Color distribution features are obtained by statistically analyzing the probability distribution of hue, saturation, and brightness in the HSV color space. The parameter vector dimension of the four types of features is unified to 128 dimensions.

[0012] Furthermore, the parameterization process described in step 2 includes feature normalization and weight allocation, which allows children to adjust the weights of various scientific features (adjustable from 0.1 to 1.0) through the human-computer interaction module. The higher the weight, the stronger the expressiveness of the corresponding feature in the creative image.

[0013] Furthermore, the generative AI model described in step 4 is a LoRA fine-tuning model based on Stable Diffusion. The fine-tuning dataset contains 10,000+ microscopic images covering the fields of biology, chemistry, and physics. After fine-tuning, the model has a ≥90% restoration rate of scientific features, and the generated image resolution supports adjustable resolution from 512×512 to 2048×2048, with a generation time of ≤10s.

[0014] Furthermore, the secondary optimization described in step 5 supports two modes: Creative adjustments (such as modifying sketches or text prompts) and feature optimizations (such as adjusting feature weights or adding / removing feature types) are used to create a comparative image of the optimized version, helping children understand the connection between scientific features and creative expression.

[0015] This invention provides an AI-based system and method for generating creative microscopic images. It offers the following advantages: 1. This invention provides an AI-based creative generation system and method for microscopic images, which transforms the texture, symmetry, shape contour, and color distribution of microscopic images into quantifiable parameter vectors as hard constraints for AI generation, ensuring that creative images retain the authenticity of scientific features while possessing the flexibility of artistic expression, thus solving the problem of pure creative generation lacking scientific support.

[0016] 2. This invention provides an AI-based microscopic image creative generation system and method. Targeting children's cognitive level, it designs a dual input mode of "microscopic image + simple sketch and text" to lower the creative threshold. At the same time, through secondary optimization functions, it guides children to explore the relationship between scientific features and creative expression, thereby enhancing the depth of scientific observation. It breaks through the single-function limitations of traditional tools, using scientific observation as "raw material" for artistic creation. The generated images contain traceable scientific features and reflect personalized creativity, realizing a closed-loop learning process of "observation-extraction-creation-understanding" and cultivating interdisciplinary thinking.

[0017] 3. This invention provides an AI-based creative microscopic image generation system and method. Based on the LoRA lightweight fine-tuning technology of Stable Diffusion, it uses a massive microscopic image dataset to optimize the model, ensuring that the model can accurately identify and restore scientific features. The generation speed and image quality are adapted to children's real-time interactive needs. The output image is accompanied by a scientific feature annotation map, clearly showing the microscopic image features integrated in the creative image, helping children understand the connection between creation and science. Attached Figure Description

[0018] Figure 1 This is a diagram illustrating the overall architecture of the AI-based microscopic image creative generation system of the present invention. Figure 2 This is a flowchart of the AI-based microscopic image creative generation method of the present invention; Figure 3 This is a schematic diagram comparing the input microscopic image and the output creative work of this invention; Figure 4 This is a schematic diagram of the human-computer interaction interface design of the present invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0020] In the description of this invention, unless otherwise explicitly specified and limited, the terms "connected," "linked," and "fixed" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0021] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0022] In the description of this embodiment, the terms "upper," "lower," "right," etc., refer to the orientation or positional relationship shown in the accompanying drawings. They are used only for ease of description and simplification of operation, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. In addition, the terms "first" and "second" are used only for distinction in description and have no special meaning.

[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0024] like Figures 1-2 As shown, this embodiment of the invention provides an AI-based microscopic image creative generation system, including an image input module, a feature analysis module, an AI generation module, a human-computer interaction module, and an output module; The image input module includes a microscope interface unit, an image uploading unit, and a preprocessing unit; The microscope interface unit supports connection to a digital microscope via USB and wireless interface to directly acquire real-time microscopic images; The image uploading unit supports uploading local microscopic image files; The preprocessing unit uses 3×3~5×5 Gaussian filtering to remove image noise, adjusts brightness and contrast through histogram equalization, and normalizes the image size to 512×512 pixels to provide standardized input for feature extraction. The feature analysis module includes a texture extraction unit, a symmetry analysis unit, a shape contour extraction unit, a color distribution statistics unit, and a parameterization unit; The texture extraction unit uses the gray-level co-occurrence matrix (GLCM) to calculate four parameters of the image in the 0°, 45°, 90°, and 135° directions, namely energy, entropy, contrast, and correlation, to comprehensively characterize the texture features of the microscopic image. The symmetry analysis unit transforms the image to the frequency domain through Fourier transform, calculates the similarity scores of left-right, top-bottom, and central symmetry, and quantifies the symmetry features. The shape contour extraction unit uses the Canny edge detection algorithm to extract the target contour, and combines the Hu moment to calculate 7 invariant moment parameters to characterize the shape's morphological features, while outputting the contour coordinate point set. The color distribution statistics unit converts the image from RGB space to HSV space, statistically analyzes the probability distribution histograms of hue (H), saturation (S), and brightness (V), and extracts peak, mean, and variance parameters. The parameterization unit normalizes the parameters of the above four types of features to the range of 0 to 1, and splices them together to form a 128-dimensional feature control set, which supports dynamic adjustment of feature weights. The AI ​​generation module includes a model fine-tuning unit and a constraint generation unit; The model fine-tuning unit is based on the Stable Diffusion V1.5 or V2.1 framework and uses LoRA lightweight fine-tuning technology. It is trained using a dataset of 10,000+ microscopic images covering biological (e.g., cells, microorganisms), chemical (e.g., crystals, reagent reactions), and physical (microstructure) fields. The backbone network parameters are frozen, and only the low-rank matrix is ​​trained to ensure that the model's reproduction accuracy of scientific features is ≥90%. The constraint generation unit fuses the feature control set with creative control signals (such as sketch outline vectors or text keyword vectors), inputs them into the fine-tuned model, and achieves deep coupling between scientific features and creative expression through a cross-attention mechanism to generate creative images. The generation time is ≤10s, and the resolution is adjustable from 512×512 to 2048×2048. The human-computer interaction module includes a sketching unit, a text input unit, and an interaction control unit; The sketching unit provides a simple drawing interface, supports pressure-sensitive pen strokes and basic graphic tools (such as lines, circles, and triangles). After children draw, the sketch outline is extracted through threshold segmentation and morphological processing to generate outline control vectors. The text input unit supports Pinyin and simplified and traditional Chinese input, has a built-in vocabulary database for children, and uses a lightweight BERT model to parse text prompts into 256-dimensional keyword vectors. The interactive control unit supports feature weight adjustment, image resolution selection, and secondary optimization triggering, adapting to children's operating habits. The output module includes an image rendering unit, an annotation generation unit, and a storage and sharing unit; The image rendering unit performs color optimization and edge smoothing on AI-generated images to improve visual effects. The annotation generation unit can automatically generate a "scientific feature annotation map", which uses different colors to annotate the scientific features such as texture, shape and color in the creative image, and includes a brief text description; The storage and sharing unit supports saving creative images and labeled images locally in JPG, PNG, and PDF formats, and supports sharing to mobile phones and printers via Bluetooth and Wi-Fi for rapid output.

[0025] This AI-based method for generating creative microscopic images includes the following steps: Step 1: Microscopic Image Acquisition and Preprocessing Children can observe specimens (such as onion epidermal cells and salt crystals) through a connected digital microscope. The image input module can directly acquire real-time microscopic images or upload saved microscopic image files. The preprocessing unit automatically uses a 3×3 Gaussian filter to remove image noise, adjusts brightness and contrast through histogram equalization, normalizes the image to 512×512 pixels, and outputs a standardized microscopic image.

[0026] Step 2: Structural Extraction and Parameterization of Scientific Features The feature analysis module performs multi-dimensional analysis on standardized microscopic images: the texture extraction unit calculates GLCM parameters, the symmetry analysis unit evaluates symmetry similarity, the shape contour extraction unit obtains edge and Hu moment parameters, and the color distribution statistics unit outputs HSV spatial distribution parameters. The parameterization unit normalizes all feature parameters to the 0~1 range and splices them together to form a 128-dimensional feature control set. Children can adjust the weights of various features through the human-computer interaction module, such as increasing the weight of texture features and decreasing the weight of symmetry features.

[0027] Step 3: Acquiring Creative Input from Children Children can draw simple ideas through the sketching unit of the human-computer interaction module (such as incorporating the texture of onion epidermal cells into a flower pattern), or input prompts through the text input unit (such as "butterfly with cell texture"). The system extracts the outline of the sketch and generates an outline control vector; it performs semantic parsing on the text prompts and generates a keyword vector, which together form the creative control signal.

[0028] Step 4: AI Model Constraint Generation The system inputs the feature control set and creative control signals into the Stable DiffusionLoRA model, which has been fine-tuned by microscopic images; The model uses a cross-attention mechanism to integrate creative expression under the constraints of scientific features, generating personalized images that combine microscopic image features with children's creativity, with a generation time of ≤10s.

[0029] Step 5: Creative Image Output and Optimization The output module renders and generates images, automatically generates scientific feature annotation maps, and simultaneously displays creative images and annotation maps; If children are not satisfied with the result, they can modify the sketch or text prompts through the human-computer interaction module, or adjust the feature weights to trigger secondary generation until they obtain a satisfactory creative image. Finally, the creative images and annotations are saved or printed through the storage and sharing unit, completing the creation process.

[0030] Implementation Case: ① Implementation Scenarios like Figure 3 As shown, the application scenario is "Elementary School Science Class - Artistic Creative Transformation of Onion Epidermal Cells", targeting 8-year-old children, to realize the transformation of microscopic images of onion epidermal cells into personalized creative artworks.

[0031] ②System Configuration Image input module: docked with a digital microscope (1080P resolution), supporting real-time acquisition of onion epidermal cell microscopic images; Feature analysis module: Texture extraction uses GLCM (0°, 45°, 90°, 135° directions), shape contour extraction uses the Canny algorithm (threshold 0.1~0.3), and color distribution statistics are based on HSV space; Human-computer interaction module: The sketching interface supports 2B, 4B, or 6B strokes; text input supports pinyin association; feature weight adjustment uses a slider mode (0.1~1.0), such as... Figure 4 As shown; AI generation module: Based on Stable Diffusion V2.1, the LoRA fine-tuning dataset contains 5000+ plant cell microscopic images, with a default generation resolution of 1024×1024 pixels; Output module: Supports saving in JPG or PNG format, and supports direct output from a home printer.

[0032] ③ Implementation steps Step 1: Microscopic Image Acquisition and Preprocessing Children observe onion epidermal cells through a digital microscope. The image input module acquires real-time microscopic images (resolution 1920×1080), the preprocessing unit uses 3×3 Gaussian filtering for noise reduction, histogram equalization to adjust contrast, normalizes to 512×512 pixels, and outputs standardized onion epidermal cell images that clearly present the reticular texture of the cell wall and the circular outline of the cell nucleus.

[0033] Step 2: Structural Extraction and Parameterization of Scientific Features The feature analysis module extracts four types of core features: Texture features: GLCM energy 0.65, entropy 0.32, contrast 0.41, correlation 0.88; Symmetry features: left-right symmetry similarity 0.72, top-bottom symmetry similarity 0.68; Shape and contour features, Hu moment parameters + cell wall network contour coordinates; Color distribution characteristics: HSV hue peak 50°, average saturation 0.45, average brightness 0.70; After parameterization, a 128-dimensional feature control set is formed. The child adjusts the weight of the texture feature to 0.9 and the weight of the shape and contour feature to 0.8 using a slider.

[0034] Step 3: Acquiring Creative Input from Children Children draw simple butterfly outlines using a sketching unit, and the system extracts the outline control vectors. Simultaneously, the input text prompt "butterfly with onion cell texture" is semantically parsed to generate keyword vectors, forming creative control signals.

[0035] Step 4: AI Model Constraint Generation The system inputs the feature control set and creative control signal into the fine-tuned AI model. Under the constraints of the texture and shape features of onion epidermal cells, the model integrates butterfly ideas to generate a creative image with "cell wall mesh texture as wing pattern and cell nucleus shape as butterfly body". The generation time is 8 seconds.

[0036] Step 5: Creative Image Output and Optimization The output module renders the creative image and automatically generates an annotation map, using red to mark the cell texture features on the wings and blue to mark the cell nucleus shape features on the body; The child felt the butterfly wings were too pale, so the color distribution feature weight was adjusted to 0.7 through the interactive module, triggering secondary generation, and finally a satisfactory creative image was obtained, which was then saved and printed.

[0037] ④ Implementation Results Scientific observation results: To optimize their creative images, children actively and repeatedly observed the texture details and shape characteristics of onion epidermal cells, improving the level of detail in scientific observation by 70% compared to traditional classrooms; Creative output quality: The generated creative images not only fully preserve the core scientific features of onion cells, such as the reticular texture and round cell nucleus, but also perfectly integrate the butterfly concept, with a scientific feature reproduction rate of 92% and creative expression that meets children's expectations; Interactive experience: The entire operation takes ≤15 minutes, which children can complete independently. The usage rate of the secondary optimization function reaches 85%, which fully stimulates creative enthusiasm. Educational value: Through the annotated diagrams, children clearly understand the connection between creativity and science. 80% of children can accurately describe that the texture of butterfly wings comes from the cell wall of onion cells, achieving a dual improvement in scientific cognition and artistic creation.

[0038] The following points should be noted in this article: 1. The accompanying drawings of the embodiments disclosed herein only relate to the structures involved in the embodiments disclosed herein; other structures can be referred to in a general design.

[0039] 2. Where there is no conflict, the embodiments of this disclosure and the features in the embodiments can be combined with each other to obtain new embodiments.

[0040] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

Claims

1. An AI-based microscopic image creative generation system, characterized by, The system includes an image input module, a feature analysis module, an AI generation module, a human-computer interaction module, and an output module; The image input module includes a microscope interface unit, an image uploading unit, and a preprocessing unit; The microscope interface unit supports connection to a digital microscope via USB and wireless interface to directly acquire real-time microscopic images; The image uploading unit supports uploading local microscopic image files; The preprocessing unit uses 3×3~5×5 Gaussian filtering to remove image noise, adjusts brightness and contrast through histogram equalization, and normalizes the image size to 512×512 pixels to provide standardized input for feature extraction. The feature analysis module includes a texture extraction unit, a symmetry analysis unit, a shape contour extraction unit, a color distribution statistics unit, and a parameterization unit; The texture extraction unit uses the gray-level co-occurrence matrix (GLCM) to calculate four parameters of the image in the 0°, 45°, 90°, and 135° directions, namely energy, entropy, contrast, and correlation, to comprehensively characterize the texture features of the microscopic image. The symmetry analysis unit transforms the image to the frequency domain through Fourier transform, calculates the similarity scores of left-right, top-bottom, and central symmetry, and quantifies the symmetry features. The shape contour extraction unit uses the Canny edge detection algorithm to extract the target contour, and combines the Hu moment to calculate 7 invariant moment parameters to characterize the shape's morphological features, while outputting the contour coordinate point set. The color distribution statistics unit converts the image from RGB space to HSV space, statistically analyzes the probability distribution histograms of hue (H), saturation (S), and brightness (V), and extracts peak, mean, and variance parameters. The parameterization unit normalizes the parameters of the above four types of features to the range of 0 to 1, and splices them together to form a 128-dimensional feature control set, which supports dynamic adjustment of feature weights. The AI ​​generation module includes a model fine-tuning unit and a constraint generation unit; The model fine-tuning unit is based on the Stable Diffusion V1.5 or V2.1 framework, employs LoRA lightweight fine-tuning technology, and is trained using a dataset of 10,000+ microscopic images covering the fields of biology, chemistry, and physics. The backbone network parameters are frozen, and only the low-rank matrix is ​​trained to ensure that the model's reproduction accuracy of scientific features is ≥90%. The constraint generation unit fuses the feature control set with the creative control signal, inputs it into the fine-tuned model, and achieves deep coupling between scientific features and creative expression through the cross-attention mechanism to generate creative images. The generation time is ≤10s, and the resolution is adjustable from 512×512 to 2048×2048. The human-computer interaction module includes a sketching unit, a text input unit, and an interaction control unit; The sketching unit provides a simple drawing interface, supports pressure-sensitive pen strokes and basic graphic tools. After children draw, the sketch outline is extracted through threshold segmentation and morphological processing to generate outline control vectors. The text input unit supports Pinyin and simplified and traditional Chinese input, has a built-in vocabulary database for children, and uses a lightweight BERT model to parse text prompts into 256-dimensional keyword vectors. The interactive control unit supports feature weight adjustment, image resolution selection, and secondary optimization triggering, adapting to children's operating habits. The output module includes an image rendering unit, an annotation generation unit, and a storage and sharing unit; The image rendering unit performs color optimization and edge smoothing on AI-generated images to improve visual effects. The annotation generation unit can automatically generate a "scientific feature annotation map", which uses different colors to annotate the scientific features such as texture, shape and color in the creative image, and includes a brief text description; The storage and sharing unit supports saving creative images and labeled images locally in JPG, PNG, and PDF formats, and supports sharing to mobile phones and printers via Bluetooth and Wi-Fi for rapid output. 2.The AI-based microscopic image creative generation system of claim 1, wherein, The image input module supports direct acquisition of microscopic images and local image upload. Preprocessing uses Gaussian filtering for noise reduction, and normalization maps image pixel values ​​to the 0~1 range to ensure the consistency of image brightness and contrast. 3.The AI-based microscopic image creative generation system of claim 1, wherein, The output module supports local saving, printing, and sharing of creative images, and generates a "scientific feature annotation map" to annotate the scientific features of the microscopic images integrated into the creative images, enabling scientific tracing of the creative process.

4. An AI-based microscopic image creative generation method, characterized by, Includes the following steps: Step 1: Microscopic Image Acquisition and Preprocessing The image input module acquires real microscopic images, and the images are preprocessed by noise reduction and normalization to ensure the accuracy of scientific feature extraction. Step 2: Structural Extraction and Parameterization of Scientific Features The feature analysis module performs structured analysis on the preprocessed microscopic images, extracting four core scientific features: texture, symmetry, shape contour, and color distribution. Each feature is then converted into a quantifiable parameter vector to form a feature control set. Step 3: Acquiring Creative Input from Children The human-computer interaction module receives simple sketches or text prompts from children, extracts the outlines of the sketches, performs semantic analysis on the text prompts, and generates creative control signals. Step 4: AI Model Constraint Generation The feature control set and creative control signal are input into a generative AI model finely tuned by a microscopic image dataset. The model integrates creative expression under scientific feature constraints to generate creative images. Step 5: Creative Image Output and Optimization The output module outputs creative images, and children can adjust the creative input or feature weights through the human-computer interaction module to achieve personalized optimization. 5.The AI-based microscopic image creative generation method of claim 4, wherein, The algorithm for extracting scientific features in step 2 is as follows: Texture features are extracted using the Gray-Level Co-occurrence Matrix (GLCM) to obtain energy, entropy, and contrast parameters. Symmetry features are analyzed using Fourier transform to determine the similarity between left-right and top-bottom symmetries of images. The shape contour features are extracted using Hu moments and the Canny algorithm for edge detection to extract contour coordinates and morphological parameters; Color distribution features are obtained by statistically analyzing the probability distribution of hue, saturation, and brightness in the HSV color space. The parameter vector dimension of the four types of features is unified to 128 dimensions. 6.The AI-based microscopic image creative generation method of claim 4, wherein, The parameterization process described in step 2 includes feature normalization and weight allocation, which allows children to adjust the weights of various scientific features through the human-computer interaction module. The higher the weight, the stronger the expressiveness of the corresponding feature in creative images. 7.The AI-based microscopic image creative generation method of claim 4, wherein, The generative AI model described in step 4 is a LoRA fine-tuning model based on Stable Diffusion. The fine-tuning dataset contains 10,000+ microscopic images covering the fields of biology, chemistry, and physics. After fine-tuning, the model has a ≥90% accuracy in restoring scientific features. The generated image resolution is adjustable from 512×512 to 2048×2048, and the generation time is ≤10 seconds. 8.The AI-based microscopic image creative generation method of claim 4, wherein, The secondary optimization described in step 5 supports two modes: Creative adjustments and feature optimizations are made, and the optimized generated images are compared with the original versions to help children understand the connection between scientific features and creative expression.

Images