A method and system for making smart ar stickers

Abstract

The application relates to an intelligent AR sticker making system and method and belongs to the technical field of augmented reality. The system realizes user registration and sticker template PNG background picture configuration through a management module; an AR function module sets a white edge parameter of 15 pixels based on an object recognition unit to realize accurate positioning; a PC beautifying module adopts a bilateral filtering algorithm with a spatial domain standard deviation sigma d = 15 pixels and a color domain standard deviation sigma r = 0.3xs v Dynamic real-time beautifying is adjusted, and the frame rate is greater than or equal to 30 fps under CUDA acceleration; a cutting machine instruction format is controlled by a CRC32 check protocol, the instruction format is an instruction header, a length, data and a CRC32 check code, and a cutting error is less than or equal to 1 mm; a recall module supports two-dimensional code sharing of electronic stickers. The system realizes full-process automation, a sticker registration error is less than or equal to 0.1 pixels, and a dynamic tracking displacement error is less than or equal to 1 cm. The application realizes integrated control of AR sticker creation, beautifying parameter optimization and printing calibration through a centralized management module, effectively solves the problem of AR special effect and printing output color distortion, and supports efficient operation of multiple devices in a commercial scene.

Description

Technical Field

[0001] This invention relates to the field of augmented reality (AR) image processing technology, and in particular to a comprehensive control system and method that integrates AR sticker creation, intelligent beautification optimization, and printing output. Background Technology

[0002] Existing AR photo-taking systems have certain limitations. In commercial scenarios, multiple devices require independent sticker configuration, leading to high maintenance costs. Mismatches between screen display color gamut and printed output cause image distortion. Users also need to perform additional operations to transfer photos. In summary, the issues of multi-device collaboration and output calibration remain unresolved. Therefore, there is an urgent need to propose a method and system for creating smart AR stickers. Summary of the Invention

[0003] To address the problems existing in current technologies, this invention provides a method and system for creating smart AR stickers. This system integrates AR sticker creation, intelligent beautification shooting, and photo printing output, overcoming the shortcomings of traditional sticker systems in terms of customizability, interactivity, and output efficiency. The system supports collaborative work between the management and user ends, achieving full automation from sticker configuration to final printing.

[0004] This invention is implemented as follows: a method and system for creating smart AR stickers, wherein the AR-based sticker processing system includes:

[0005] Management module: Includes information management unit, sticker collection unit, and printing unit; used for user information management, including registration, modification, sticker collection, and print output management;

[0006] AR Function Module: Includes object recognition and tracking unit, AR screenshot unit, and sticker unit; used for AR recognition, custom AR stickers and sending, saving and sharing AR images, and viewing AR stickers imported from the module; the sticker unit has the function of generating and publishing custom AR commemorative stickers;

[0007] PC beautification module: Includes beautification parameter settings, beautification presets, and mirror mode; used for real-time image enhancement to improve user experience.

[0008] The "Recollection" module includes sections for recommending, publishing, and viewing reflections; it facilitates sharing and interaction among users through images and short articles.

[0009] Furthermore, the management module specifically includes:

[0010] Information Management Unit: Responsible for user registration, login, and account management. Users register by entering their name and mobile phone number. After the administrator approves and activates the registration in the backend, the user can log in. After logging in, the user is taken to the sticker management page, where account information can be modified.

[0011] Sticker Collection Unit: Allows administrators to add, modify, and hide sticker categories and sticker effects on the management interface. Users can collect frequently used sticker templates for quick access. Supports category management (adding, hiding, and deleting categories) and sticker management (modifying DIY areas and updating items).

[0012] Printing Unit: Manages the print output process, including device calibration and print configuration. Image position, scaling, marker size and position are set via the print management page to ensure print accuracy. One-click printing is supported, and photos are automatically cropped after printing.

[0013] Furthermore, the AR function module specifically includes:

[0014] Object recognition and tracking unit: Based on real-time camera recognition of human outlines, it achieves accurate tracking in conjunction with a 1080P or higher resolution camera. In the DIY area, the default white border parameter for the human figure is set to 15 to ensure white space is left when cutting the sticker.

[0015] Sticker section: Supports custom AR sticker creation, including settings for background image, foreground image, shooting area, and AR style. Background images are PNG format, 1200x1800 pixels, displayed behind the subject. Foreground images can be moved, scaled, copied, or deleted. AR style frames are dragged and dropped onto the shooting frame to generate corresponding effects. Stickers can be linked as commemorative stickers and support digital download and sharing.

[0016] AR Screenshot Unit: Automatically generates AR effect images during the user's shooting process. Supports automatic shooting with a 5-second countdown. After shooting, a confirmation page is displayed, where users can choose to retake the photo or immediately generate an AR sticker.

[0017] Furthermore, the PC beautification module specifically includes:

[0018] Beauty settings: Adjust the parameters using the slider, ranging from 0 to 1, including skin smoothing, contour optimization, etc. Click the "Apply" button to save after setting.

[0019] Beauty presets: Offers 5 preset modes: Natural, Refined, Influencer, 3D, and Cute. Clicking the corresponding button automatically sets the parameters; a reset button is also supported to reset the settings to zero.

[0020] Mirror Mode: When checked, the shooting screen will be flipped horizontally to suit selfie habits.

[0021] Furthermore, the memory module specifically includes:

[0022] The "Remembering Memories" section intelligently recommends relevant memories based on users' sticker collections and shooting history. Users can preview the photos on the app and access digital copies via a QR code for easy sharing.

[0023] The "Reflections" section allows users to post sticker photos and short reflections. After clicking the download QR code on the preview page, users can save the image and share it on social media.

[0024] The "Thoughts & Feelings Viewing" section allows users to view AR stickers and their thoughts shared by other users, creating an interactive community. The user homepage and sticker selection page display popular templates, facilitating content discovery.

[0025] In summary, the advantages and positive effects of this invention are as follows:

[0026] This invention constructs a complete intelligent AR sticker creation ecosystem through deep collaboration of management modules, AR function modules, PC beautification modules, and memory modules. Its core breakthrough lies in integrating the discrete steps of traditional sticker production—design, shooting, output, and sharing—into a fully automated process, solving three major industry pain points: poor customization, weak interactivity, and low output efficiency. The system adopts a modular design to achieve a closed-loop technology: after user registration and administrator approval to generate a unique ID account, the user can configure sticker templates on the management end—setting a 1200×1800 pixel PNG background image, a draggable and zoomable foreground image, and a precisely bound AR style frame to the shooting area through the DIY area; after the user selects a template, a 5-second countdown automatic focus intelligent shooting is triggered, combined with real-time rendering of the PC beautification module with 5 preset modes applied with one click to generate high-quality AR stickers; finally, the output is calibrated by printing units automatically cut by a dye-sublimation printer, and the memory module distributes the stickers as electronic QR codes for sharing, forming a complete value chain from creation to dissemination.

[0027] Because AR-based sticker processing systems incorporate both virtual and real-world elements, they create a highly immersive environment. Augmented reality systems are emerging as a new medium, with tracking registration technology being one of the most crucial technologies. Registration is essentially the process of "aligning" computer-generated virtual objects with scenes in the real environment. Registration must first determine the relationship between the virtual object and the observer, then project the virtual object outside the observer's field of view using correct geometric projection—one method being affine transformation. AR-based sticker information processing systems, with tracking registration technology at their core, achieve an immersive experience through precise "alignment" between the virtual and real environments. The core of registration technology lies in establishing the geometric mapping relationship between virtual objects and real scenes; specific methods and technical support are as follows:

[0028] Beauty algorithm formula

[0029] The beautification algorithm is implemented in the PC beautification module, employing a bilateral filtering algorithm for real-time skin smoothing. The core algorithm is based on a dual filtering mechanism in both the spatial and color domains, and runs on the NVIDIA Jetson Nano platform with CUDA acceleration, ensuring a frame rate of ≥30fps. The complete derivation of the formula and parameter explanations are as follows: 1. Formula for bilateral filtering algorithm Bilateral filtering smooths images by weighted averaging of pixel values ​​while preserving edge details. The formula is defined as follows: Where I(x+i,y+j) are the pixel values ​​of the input image. kernel(i,j) is the bilateral filter kernel function, consisting of spatial domain weights and color domain weights, as kernel(i,j)=w d (i,j)·w r (i,j); w d (i,j) represents the spatial domain weights, calculated based on pixel distance: w r (i,j) color gamut weights, calculated based on pixel intensity differences: 2. Dynamic adjustment of core parameters The parameter is based on the user's slider value (s) v The normalized interval (0-1) is adjusted in real time. Spatial domain standard deviation σ d : Controls the smoothing range; default value 15 pixels (adjustable range 10-20 pixels). Formula: σ d =15 indicates that a fixed value of 15 ensures smooth skin areas while avoiding excessively blurred edges (such as the hairline). Color gamut standard deviation σ r To prevent color distortion, σ is dynamically calculated based on the slider value. r =0.3×s v This indicates that when the slider value s v When σ = 0, r =0 (no beautification); s v When σ = 1, r =0.3 (maximum beautification intensity), balancing skin tone evenness and naturalness. 3. Implementation details and performance optimization Hardware acceleration: CUDA parallel computing is used and implemented on the NVIDIA Jetson Nano platform; when processing 1080P images, the latency is ≤33ms (corresponding to a frame rate ≥30fps). Kernel function optimization: Use shared memory to reduce global access and improve computational efficiency. Real-time rendering process: Users set the s via a slider vValue (0~1); the system calculates σ r =0.3×s v The system applies a bilateral filtering kernel function to output an enhanced image; and combines facial landmark detection (using the Dlib library) to locally optimize contour regions (such as cheeks and chin). 4. Application Examples Assuming the user sets the slider value s v =0.5:σ r =0.3×0.5=0.15, the filtering kernel focuses on smoothing skin texture (σ d =15 fixed), while preserving the edges of facial features. Output effect: Natural skin tone transition, without any "plastic" distortion. 5. Advantages of the beautification algorithm Adjusting the dynamic parameter (σ) r =0.3×s v This allows for personalized customization while maintaining real-time performance and quality (frame rate ≥ 30fps).

[0030] Communication protocol between printing unit and die-cutting machine

[0031] The printing unit manages the output process, communicating with the die-cutting machine via a dedicated protocol to ensure printing accuracy and automatic cutting functionality. The protocol is designed for low latency and high reliability, supporting one-click printing. The following is a complete description of the protocol: 1. Core Structure of the Protocol The protocol uses a binary format and consists of three parts: initialization handshake, instruction transmission, and error handling. (1) Initial handshake: Establish device connection. Send command: 'INIT\xFF\x00' (hexadecimal: 494E4954FF00) Expected response: 'ACK\xFF\x01' (hexadecimal: 41434BFF01) Timeout mechanism: If there is no response within 3 seconds, retry will be performed, up to a maximum of 2 times. (2) G command transmission format: used to send print and cut commands. General format: [Command Header] + [Length] + [Data] + [CRC32 Checksum] Instruction header: 1 byte, identifies the instruction type (e.g., G indicates a general instruction). Length: 2 bytes, indicating the number of bytes in the data field. Data: Variable length, containing specific parameters (such as coordinates, power). CRC32 checksum: 4 bytes, calculated based on IEEE standard, to ensure data integrity. (3) Error retransmission mechanism: triggered when verification fails. Maximum number of retransmissions: 3. Error handling process: Verification failure → Send NAK signal → Retransmit original command. 2. Examples of key instructions (1) Print command: Send image data and print parameters. Command header: 'P' (print) Data fields include image location (X,Y coordinates), scaling ratio (floating-point number), and marker size (pixels). Example: 'P' + [length = 12] + [X = 100, Y = 200, scaling = 1.0] + [CRC32] (2) Cutting command: Trigger the die-cutting machine to cut automatically. Command header: 'C' (cut) Data field: contains the coordinates of the cutting path. Example: 'C' + [length = 8] + [path data] + [CRC32] (3) Calibration command: Executed when the device is started. Command header: 'G' (General) Data field: Zeroing instruction G28X0Y0 (reset coordinates). 3. Error handling mechanism (1) Error code definition (2) Retransmission logic First failure: Retransmit immediately. Second failure: Retransmission after a 500ms delay. Third failure: Terminate the process and log the error. 4. Three-level fault tolerance mechanism Initial verification failed: Retransmit command immediately Second failure: 500ms delay for retransmission + device self-test initiated Three failures: The process terminates and reports error code ERR03 (system-level failure). 5. Advantages of communication protocols High reliability design (CRC32 check + error retransmission) ensures printing and cutting accuracy (error ≤1mm) and supports one-click automated output. 6. System Coordination The beautification module outputs images directly into the printing unit, forming a closed-loop process (user-end shooting → beautification processing → protocol transmission → printing and cutting).

[0032] Feature-based registration method: This method extracts feature points from the real scene, such as the outline of a person and background texture, using the SIFT algorithm, and matches them with predefined feature points of the virtual sticker. A homography matrix is ​​calculated to establish the mapping relationship, as shown in the following formula: Where (x,y) are the coordinates of the virtual sticker, (x′,y′) are the coordinates projected onto the real scene, and H is the homography matrix. H is solved by the RANSAC algorithm with ≥100 iterations.

[0033] The system application of the feature-based registration method is as follows: When the user drags the AR style frame to the photo area, the object recognition unit extracts human contour feature points such as shoulder lines and hairlines in real time; the system matches the corner points of the style frame with human feature points, and accurately superimposes the virtual effects onto the top of the person's head through affine transformation with a position error of ≤0.1 pixels.

[0034] Model-based registration method: A pre-constructed 3D point cloud model of the real scene is acquired via an RGB-D camera, and the 3D model of the virtual sticker is aligned with the scene point cloud using the ICP (Iterative Closest Point) algorithm. Transformation formula: Where p i For scene point clouds, q i Let T be the virtual model point cloud, and T be the rigid body transformation matrix.

[0035] The system application of the model-based registration method is as follows: in a fixed scene such as a shopping mall check-in point, the system calls the pre-stored scene point cloud model. Virtual landmark stickers are bound to specific locations in the real scene, such as the center of a wall, using an ICP algorithm to ensure that the sticker's position remains stable when the viewing angle changes.

[0036] Region-based registration method: Define the ROI (Region of Interest), and locate the adaptation range of the virtual sticker through edge detection (Canny operator) and texture analysis (LBP features). Position adjustment formula: E is the similarity energy function between the ROI region and the virtual sticker, and k is the step size coefficient.

[0037] System Application: When the "People Background" template is selected, the photo area (semi-transparent frame) is used as the ROI; the background image is automatically filled into the ROI area, and the position is optimized by gradient descent to achieve seamless integration of the people and background edges.

[0038] Real-time dynamic registration based on SLAM: Real-time estimation of camera pose T using the ORB-SLAM2 framework. cw(From world coordinate system to camera coordinate system), the world coordinates P of the virtual sticker w Project onto the image plane using the following formula: P c =T cw ·P w p = K·P c (5) K is the camera intrinsic parameter matrix, and p is the 2D pixel coordinates after projection.

[0039] System Applications: During user-controlled shooting, the SLAM thread continuously updates the camera pose; the virtual pet sticker's world coordinates P... w It is attached to the character's shoulder and the follow effect is kept with a jitter error of less than 2 pixels through real-time projection.

[0040] The technical support and system integration include hardware collaboration, algorithm optimization, and calibration mechanisms.

[0041] Hardware collaboration includes depth cameras and IMU sensors.

[0042] Depth camera: Intel RealSense D415 provides scene point cloud data (for model registration);

[0043] IMU sensor: Integrated into touch screen devices to assist in SLAM pose estimation (angular velocity error <0.5° / s).

[0044] Algorithm optimization includes accelerating feature matching and ensuring real-time performance.

[0045] Accelerated Feature Matching: ORB features are used instead of SIFT (10x speedup in computation);

[0046] Real-time performance guarantee: SLAM and registration algorithms are computed in parallel on the GPU (NVIDIA Jetson Nano), with a frame rate of ≥30fps.

[0047] The calibration mechanism includes offline calibration and online correction.

[0048] Offline calibration: Camera intrinsic parameter K is calibrated using a checkerboard pattern and stored in the system configuration file;

[0049] Online calibration: The internal parameter verification is automatically run once every 24 hours, and K is calculated inversely using environmental feature points.

[0050] Through the above methods, the system achieves geometric alignment and dynamic adaptation between virtual stickers and the real environment. Specifically, in static scenes, based on feature registration technology, the positioning accuracy between the virtual sticker and the real environment reaches sub-pixel level, with a feature registration error ≤0.1 pixels (at 1080P resolution), ensuring natural blending of the sticker and background. For dynamic scenes, relying on SLAM (Simultaneous Localization and Mapping) technology, the system can track camera pose in real time and dynamically adjust sticker projection parameters. The SLAM tracking displacement error is controlled within 1cm (within 1m), ensuring stable adhesion to the real environment even when the user moves or the scene changes slightly. Furthermore, the system adopts a unified algorithm framework, with core functions implemented based on the OpenCV 4.5 algorithm library across different operating systems such as Windows, Android, and iOS, ensuring consistency in cross-platform development and deployment and lowering the technical threshold for multi-device adaptation. These technical indicators collectively guarantee the immersiveness and reliability of the system in practical applications. Whether in static displays or dynamic interactive scenes, the virtual sticker can seamlessly blend with the real environment, providing users with a high-quality AR experience. Attached Figure Description

[0051] Figure 1: System Module Architecture Diagram

[0052] Figure 2: User Registration and Review Flowchart

[0053] Figure 3: Basic information and DIY area diagram of the sticker set

[0054] Figure 4: Background configuration interface diagram

[0055] Figure 5: AR sticker shooting and production process flowchart

[0056] Figure 6: Beauty Parameter Settings Panel Figure 7: Printing Calibration and Output Flowchart Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0058] The application principle of the present invention will be further described below with reference to the accompanying drawings.

[0059] like Figure 1 As shown, the AR-based sticker processing system provided in this embodiment of the invention includes: a management module, an AR function module, a PC beautification module, and a memory module;

[0060] Management module: Includes information management unit, sticker collection unit, and printing unit; used for user information management, including registration, modification, sticker collection, and print output management;

[0061] AR Function Module: Includes object recognition and tracking unit, AR screenshot unit, and sticker unit; used for AR recognition, custom AR stickers and sending, saving and sharing AR images, and viewing AR stickers imported from the module; the sticker unit has the function of generating and publishing custom AR commemorative stickers;

[0062] PC beautification module: Includes beautification parameter settings, beautification presets, and mirror mode; used to beautify captured images in real time and improve user experience;

[0063] The "Recollection" module includes sections for recommending, publishing, and viewing reflections; it facilitates sharing and interaction among users through images and short articles.

[0064] Furthermore, the management module includes: an information management unit, a sticker collection unit, and a printing unit;

[0065] The information management unit is responsible for user registration, login, and account management. Users register by entering their name and mobile phone number. After the administrator reviews and activates the registration in the backend, the user can log in. After logging in, the user is taken to the sticker management page, where account information can be modified.

[0066] Sticker Collection Unit: Allows administrators to add, modify, and hide sticker categories and sticker effects on the management interface. Users can collect frequently used sticker templates for quick access. Supports category management (adding, hiding, and deleting categories) and sticker management (modifying DIY areas and updating items).

[0067] Printing Unit: Manages the print output process, including device calibration and print configuration. The print management page allows you to set image position, zoom level, marker size and position to ensure print accuracy. It supports one-click printing and automatically crops photos after printing.

[0068] Furthermore, the AR function module specifically includes:

[0069] Object recognition and tracking unit: Based on real-time camera recognition of human outlines, it achieves accurate tracking in conjunction with a 1080P or higher resolution camera. In the DIY area, the default white border parameter for the human figure is set to 15 to ensure white space is left when cutting the sticker.

[0070] Sticker section: Supports custom AR sticker creation, including settings for background image, foreground image, shooting area, and AR style. Background images are PNG format, 1200x1800 pixels, displayed behind the subject. Foreground images can be moved, scaled, copied, or deleted. AR style frames are dragged and dropped onto the shooting frame to generate corresponding effects. Stickers can be linked as commemorative stickers and support digital download and sharing.

[0071] AR Screenshot Unit: Automatically generates AR effect images during the user's shooting process. Supports automatic shooting with a 5-second countdown. After shooting, a confirmation page is displayed, where users can choose to retake the photo or immediately generate an AR sticker.

[0072] Furthermore, the PC beautification module specifically includes:

[0073] The beauty filter settings use a bilateral filtering algorithm and a CUDA acceleration scheme.

[0074] Beauty presets: Offers 5 preset modes: Natural, Refined, Influencer, 3D, and Cute. Clicking the corresponding button automatically sets the parameters; a reset button is also supported to reset the settings to zero.

[0075] Mirror Mode: When checked, the shooting screen will be flipped horizontally to suit selfie habits.

[0076] Furthermore, the memory module specifically includes:

[0077] The "Remembering Memories" section intelligently recommends relevant memories based on users' sticker collections and shooting history. Users can preview the photos on the app and access digital copies via a QR code for easy sharing.

[0078] The "Reflections" section allows users to post sticker photos and short reflections. After clicking the download QR code on the preview page, users can save the image and share it on social media.

[0079] The "Thoughts & Feelings Viewing" section allows users to view AR stickers and their thoughts shared by other users, creating an interactive community. The user homepage and sticker selection page display popular templates, facilitating content discovery.

[0080] The following is combined with Figures 2-6This embodiment provides a detailed description of the method and workflow for creating smart AR stickers. First, a portrait photo is taken using a camera. This step can be done anywhere, as long as the lighting in the shooting environment is sufficient and the portrait is clear. Then, a preferred PC beautification or automatic beautification method is selected. After shooting, the portrait photo is converted into an image with a specific artistic style using a cloud-based AR rendering algorithm. The AR rendering algorithm can employ deep learning-based neural network style transfer technology, supporting the conversion of multiple artistic styles and allowing for customization according to user needs. After rendering, the software automatically sends the image to a connected sticker printing device for printing. The printing device automatically adjusts printing parameters based on the image size and quality to ensure excellent sticker quality. After printing, the sticker is automatically fed into a cutting machine via an automatic conveyor for trimming. The cutting machine automatically cuts the sticker according to the image edges.

[0081] The system uses user interaction input as parameters for AR rendering, then collects real-time photo data from the user, embeds it into diverse photo templates for layout, and processes the image using algorithms. Next, user interaction is used again for secondary preview template selection, frame selection, and signature customization, before finally performing local compositing, printing, uploading, and cloud AR compositing.

[0082] The system sends a thumbnail of the photo to the cloud, and then the cloud calls the AR rendering parameters through the API, processes the image for AR rendering, and returns it to the local machine.

[0083] The method and system for creating smart AR stickers in this embodiment include the following steps:

[0084] Step 1: User registration and account activation is the information management unit of the management module.

[0085] Front-end input: Users enter their name and mobile phone number via touchscreen and click the [Register] button;

[0086] Backend verification: The system sends a mobile number deduplication request to the database. If the number is not registered, it returns a "registerable" status.

[0087] Information submission: Users can fill in basic information such as age and occupation (optional), and the system stores the password using MD5 encryption.

[0088] Administrator approval: If the status of the account to be activated is pending when it is generated in the backend, the administrator will activate the account after approval in the management terminal;

[0089] ID generation: The system assigns a unique user ID in the format: USER_date_random code, which is used for authentication in subsequent operations;

[0090] User interface: Registration and login are implemented through a unified form. After logging in, you will be redirected to the sticker management page.

[0091] Step 2: Configure the sticker template as a sticker collection unit in the management module.

[0092] Category creation: The administrator clicks the [Add Category] button, enters a category name such as "Holiday Theme", and saves.

[0093] The DIY area configuration is divided into background image, foreground image, photo area, and AR style.

[0094] Background image: Upload a 1200×1800 pixel PNG image, supporting 4 background designs;

[0095] Foreground image: Add a PNG layer, and adjust the position and size of the blue dot by dragging / scaling it;

[0096] Photo area: Set a semi-transparent frame to position the portrait; the frame will not be displayed in the actual print.

[0097] AR Style: Drag the style frame to the photo frame to bind special effects.

[0098] Parameter settings: The default width of the white border for the character is 15 pixels to ensure sufficient white space during cutting;

[0099] Save and publish: Click the [Save Project] button to update the template library on the user's end.

[0100] Key technical points: The sticker set information and DIY configuration are associated with the database field template_config.

[0101] Step 3: The entire process of shooting AR stickers and generating them into AR function modules.

[0102] User-side operation: Click [Start Creating] → Select a template → Click [Start Taking a Photo] to trigger a 5-second countdown; the camera automatically focuses on human contour tracking based on the object recognition unit;

[0103] AR effect overlay: The system renders dynamic effects using OpenGL based on preset AR style box coordinates; and generates composite images with background / foreground layers in real time.

[0104] Shooting Confirmation: The user selects [Reshoot] or [Generate Now]; after generation, the user is redirected to a preview page, which displays an electronic sticker with a QR code.

[0105] Data Stream: The AR screenshot unit stores the synthesized image as a JPG file, associated with the user ID and template ID.

[0106] Step 4: The beautification process is rendered in real time by the PC beautification module.

[0107] Dynamic parameter adjustment: The management terminal presets 5 beauty modes: Natural, Refined, Influencer, 3D, and Cute, with parameter ranges from 0 to 1; when the user takes a picture, the beauty algorithm is applied in real time, mainly based on GPU-accelerated skin smoothing processing;

[0108] Mirror mode: When checked, the screen will be flipped horizontally to suit selfie scenarios;

[0109] Synchronized effect: After modifying the parameters, click the [Apply] button to immediately apply the changes to the user's camera feed.

[0110] The core of the algorithm is to use the Dlib library to detect facial landmarks and achieve local beautification.

[0111] Step 5: Print the output as the management module's print unit.

[0112] Equipment calibration: Adjust the image position / scaling ratio in the print management page to control the error within ±1mm; set the size and position of the marker points to ensure cutting accuracy;

[0113] One-click printing: The user clicks the [One-click Print] button on the preview page; the dye-sublimation printer outputs the photo, and the die-cutting machine automatically cuts it according to the outline;

[0114] Digital distribution: Scan the QR code to download the digital stickers. The link is valid for 72 hours.

[0115] Hardware requirements: Connects to printing equipment via a high-speed interface and supports high-resolution display.

[0116] Step 6: Social interaction forms a closed loop for the memory module.

[0117] Content Recommendations: The "Thoughts & Feelings Recommendation" section analyzes users' sticker collection history and pushes similar themed stories; the user's homepage displays popular templates (stickers with more than 100 likes);

[0118] Story posting: Users scan the electronic QR code → jump to the posting page; upload sticker photos + short text (≤200 words), and associate with geolocation tags;

[0119] Community interaction: Other users can like / leave comments on the story details page; the "Thoughts and Feelings" section aggregates UGC content to form a "memory wall".

[0120] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A system for creating smart AR stickers, characterized in that... include: (1). Management module (100), used for user information management and print output control, including: The information management unit (101) performs user registration, login, and account information modification. The sticker collection unit (102) allows administrators to configure sticker categories and effect templates; The printing unit (103) controls the die-cutting machine through the CRC32 verification protocol. The instruction format is [instruction header] + [length] + [data] + [CRC32 checksum], and the cutting error is ≤1mm. The offline calibration unit (104) stores the camera intrinsic parameter matrix K; (2). AR function module (200), including: The object recognition and tracking unit (201) recognizes the outline of a person based on a camera with a resolution of 1080P or higher, and sets the white border parameter to 15 pixels. Sticker unit (202) with a PNG background image (1200×1800 pixels) and a draggable AR-style frame; AR screenshot unit (203) triggers a 5-second countdown for automatic shooting; (3). PC beauty module (300), uses bilateral filtering algorithm for real-time rendering, spatial domain standard deviation σ d =15 pixels, color gamut standard deviation σ r =0.3×s v (s v (User slider value 0-1), CUDA acceleration achieves a frame rate ≥30fps; (4). Memory module (400): Generates electronic sticker download QR code, supporting users to publish graphic and text content.

2. The system according to claim 1, characterized in that: The sticker collection unit (102) allows administrators to perform the following operations on the management terminal: Add / hide sticker categories; Configure the background image, foreground image, and photo area for the DIY area; Bind an AR-style frame to the photo frame to generate special effects.

3. The system according to claim 1, characterized in that: The communication protocol of the printing unit (103) includes: Initialize the handshake command INIT / xFF / x00 and the response ACK / xFF / x01; The print command format is P+[length]+[X coordinate, Y coordinate, scaling ratio]+[CRC32]; Three-level fault tolerance mechanism: if the first verification fails, it will be retransmitted immediately; if the second failure fails, it will be retransmitted after a 500ms delay and start a self-check; if the third failure fails, the process will be terminated.

4. The system according to claim 1, characterized in that: The PC beautification module (300) includes: 5 preset modes (natural, delicate, internet celebrity, 3D, cute); mirror mode switch; and local contour optimization based on facial key point detection.

5. The system according to claim 1, characterized in that: The memory module (400) records recommended content through user sticker collections and displays popular templates with more than 100 likes on the user's homepage.

6. A method for creating smart AR stickers, characterized in that... Including the following steps: After registration, users are reviewed by the administrator to generate a unique ID account. Configure a PNG background image and an AR-style frame in the DIY area; Image enhancement is achieved in real time using a bilateral filtering algorithm, where σ d =15 fixed, σ r =0.3×s v Dynamic adjustment; A feature-based registration method is adopted, which extracts scene feature points through the SIFT algorithm and solves the homography matrix H through the RANSAC algorithm, with a registration error ≤ 0.1 pixels; The printing unit sends a CRC32 verification command to the die-cutting machine to achieve automatic cutting and output.

7. The method according to claim 6, characterized in that: The execution process of the bilateral filtering algorithm includes: user setting slider value s v (0-1); Calculate σ r =0.3×s v ; Application spatial domain weights and color gamut weight The output image has its facial contour region optimized using the Dlib library.

8. The method according to claim 6, characterized in that: The registration method includes: Static scenes use homography matrix projection: H is solved by ≥100 RANSAC iterations; Dynamic Scene: T cw Let P be the transformation matrix from the world to the camera. w For the virtual sticker world coordinates, the projection formula P c =T cw ·P w p = K·P c Displacement error ≤ 1cm (within 1m); 9. The method according to claim 6, characterized in that: The printing instructions include: The cutting command format is C + [length] + [cutting path data] + [CRC32]; Calibration command G28X0Y0 resets device coordinates. A computer-readable storage medium storing a computer program, characterized in that: when the program is executed by a processor, it implements the steps of the method according to any one of claims 6-9.

10. The system according to claim 1, characterized in that: The offline calibration unit (104) obtains the camera intrinsic parameter matrix K through checkerboard calibration and stores it in the system configuration file.

Images