Free product layout system using image recognition

Abstract

To provide a POS system that automatically removes the background of product images and allows for free layout. [Solution] The product layout free arrangement system comprises a photographing means for photographing products, an image processing means for recognizing product parts from the photographed image using artificial intelligence and separating them from the background, a display layout means for positioning the separated product images in any position, and a database means for associating and managing product information. The image processing means achieves highly accurate background removal through image segmentation using deep learning technology, and the display layout means allows product images to be freely positioned using intuitive drag-and-drop operations. This breaks away from the conventional fixed grid format and enables product displays with high visual appeal.

Description

[Technical Field]

[0001] The present invention relates to the processing and display of product images in a POS (Point Of Sale) system, and in particular to a system that uses image recognition technology with artificial intelligence to automatically remove backgrounds from product images and arrange them in a free layout together with product information. [Background technology]

[0002] In conventional POS systems, when registering product images, rectangular images taken with a digital camera or other device are typically used and arranged in a fixed grid layout along with product name and price information. With this method, rectangular images including the background are arranged in a regular pattern, which lacks visual appeal and makes it difficult to effectively express the product's features.

[0003] Furthermore, in conventional image processing, removing backgrounds from product images has to be done manually using image editing software, which requires a lot of time and effort. Particularly in stores that handle a large number of products, it is not practical to perform such processing on all product images. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] 2023-100807 Summary of the Invention [Problem to be solved by the invention]

[0005] The present invention has been made in consideration of the above-mentioned problems, and aims to realize visually appealing product displays by automatically removing the background using image recognition technology based on artificial intelligence when registering product images in a POS system, and allowing product images to be arranged in any layout. [Means for solving the problem]

[0006] In order to solve the above problems, the product layout free arrangement system using image recognition of the present invention is characterized by comprising a photographing means for photographing products, an image processing means for recognizing product parts from the photographed image using artificial intelligence and separating them from the background, a display layout means for placing the separated product images in any position, and a database means for associating and managing product information. [Effects of the Invention]

[0007] According to the present invention, the background removal process for product images can be automated in a POS system, significantly improving work efficiency.

[0008] In addition, product images can be freely arranged, effectively expressing the product's features and enhancing its visual appeal to customers. Furthermore, using white plates or clear backgrounds can improve the recognition accuracy of AI. DETAILED DESCRIPTION OF THE INVENTION

[0009] These and other aspects, features, and advantages of the present disclosure will become apparent from the following detailed written description of the preferred embodiments and aspects taken in conjunction with the following drawings, variations and modifications of which may be made without departing from the spirit and scope of the novel concepts of the present disclosure. Aspects of one embodiment of the present disclosure may be combined with or substituted for one or more aspects of another embodiment of the present disclosure, to the extent not inconsistent.

[0010] The following disclosure provides many different embodiments and examples for implementing different features of the presented subject matter. To simplify the disclosure, specific examples of components and arrangements are disclosed below. Of course, these are merely examples and are not intended to be limiting. For example, a structure in which a first feature is covered by or in contact with a subsequently disclosed second feature may include an embodiment in which the first and second features are formed in direct contact, as well as an embodiment in which an additional feature is formed between the first and second features to prevent direct contact between the first and second features. Furthermore, the disclosure may repeat reference numbers and / or letters in various examples. Such repetition is for the purposes of brevity and clarity and does not, in itself, require a relationship between the various embodiments and / or configurations described. Furthermore, when a first element is described as being "coupled" or "coupled" to a second element, such a description includes embodiments in which the first and second elements are directly coupled or coupled to each other, as well as embodiments in which the first and second elements are indirectly coupled or coupled to each other via one or more intervening elements. As used herein, the phrase "at least one of" encompasses all exemplary variations. For example, the phrase "comprises at least one of A, B, or C" is equivalent to "consisting of A, B, and C and combinations thereof," and encompasses all possible variations of A, B, C, A+B, A+C, B+C, and A+B+C. In this disclosure, the use of a machine, an electronic operator, or a computer may include embodiments of a method, a recording medium, an apparatus, or a program. As used herein, the statement "A is B" can be replaced with "A includes B" unless there is a contradiction or unless otherwise stated in the specification.Terms used in this disclosure, including terms in the claims, may be interpreted in light of the descriptions in the specification and, unless otherwise indicated in the claims, may be interpreted based on what one or more citizens have so called, displayed, understood, or performed, or may have so called, displayed, understood, or performed, in the past, present, or future, unless otherwise indicated in the claims. In one embodiment, the system of the present invention is integrated with a POS register system. This system comprises a camera device for photographing products, a computing unit for image processing, a display device for displaying the processed results, and a database system for storing product information. The camera device is preferably a high-resolution digital camera, but is not limited to this. Various imaging means can be used, such as a smartphone camera, a built-in camera in a tablet device, or a dedicated imaging device. Using a white or solid-color background in the imaging environment can improve the accuracy of product recognition by artificial intelligence, but is not necessarily limited to this.

[0011] In one embodiment, the image processing means implements an image segmentation algorithm using deep learning technology. This algorithm preferably employs an architecture based on a convolutional neural network (CNN), but is not limited to this. It is also possible to combine it with other machine learning methods and image processing technologies. Specifically, segmentation models such as U-Net, Mask R-CNN, and the DeepLab series can be used. Furthermore, recognition accuracy can be further improved by optimizing a pre-trained model for a specific product category using transfer learning. It is important to select or combine appropriate models, taking into account the balance between processing speed and accuracy.

[0012] In one embodiment, the system includes a preprocessing function for product images. Captured images are first resized to an appropriate size, and then, as necessary, color correction, contrast adjustment, noise removal, and other processes are applied. These preprocessing steps are important for improving the accuracy of subsequent recognition processing using artificial intelligence. Preprocessing parameters are preferably automatically adjusted based on the shooting environment and product characteristics, but can also be manually adjusted by the user. The system may also include functions for combining multiple images to generate a high dynamic range (HDR) image or integrating images captured from different angles.

[0013] In one embodiment, the background removal process combines multiple image processing techniques, including edge detection, color analysis, and texture analysis. The AI ​​model processes this information in an integrated manner to accurately identify the boundary between the product and the background. In particular, for products with transparent containers or reflective surfaces, a special processing algorithm is applied to achieve more accurate cropping. The processing results are saved in an image format with an alpha channel, allowing for flexible placement in subsequent layout processing. It can also output a score indicating the reliability of the process and provide an interface to prompt manual corrections as needed.

[0014] In one embodiment, the layout arrangement function provides an intuitive operation interface using drag and drop. Users can place cropped product images anywhere on the screen and freely change their size, rotate, and adjust the stacking order. Text information such as product name, price, and description can be associated with the placed image, and this information can also be placed anywhere. The layout template function allows frequently used layout patterns to be saved and reused. It also includes a function to apply various design elements and effects to easily create special layouts tailored to the season or campaign.

[0015] As one embodiment, this system can employ a cloud-based architecture. By executing the computationally intensive parts of image processing on a cloud server, high-speed processing is achieved without relying on the processing power of the terminal. Major cloud services such as Amazon Web Services, Google Cloud Platform, and Microsoft Azure can be used, but processing can also be configured to be performed on an on-premises server. Secure and efficient data transfer is achieved by using a RESTful API or WebSocket over HTTPS as a communication protocol. Furthermore, a function for adaptively switching between local and cloud processing can be implemented to accommodate use in environments with unstable network connections.

[0016] In one embodiment, the database system manages product images, metadata, layout information, and other information in an integrated manner. Relational databases (MySQL (registered trademark), PostgreSQL, etc.) or NoSQL databases (MongoDB, DynamoDB, etc.) can be used, with an appropriate database system selected depending on the characteristics and scale of the data. Product images are stored in an object storage service, and only reference information is stored in the database, achieving efficient data management. Version control functionality makes it possible to track changes to product information and layouts and restore them to previous states as necessary. Furthermore, when sharing data between multiple stores, an appropriate synchronization mechanism is implemented.

[0017] In one embodiment, the system is equipped with a multilingual function. Text information such as product names and descriptions can be registered in multiple languages, and the appropriate language is automatically selected when displayed. Language switching is based on system settings or user selection. By linking with an automatic translation API, it is also possible to support unregistered languages. UTF-8 is used as the standard character encoding, and various languages ​​around the world are supported. Layout processing provides an adjustment function that takes into account differences in character display direction (e.g., left-to-right, right-to-left) and character size, which vary depending on the language.

[0018] In one embodiment, security features such as user authentication, access control, and data encryption are implemented. User authentication can be achieved using methods such as password authentication, biometric authentication, and multi-factor authentication. Access control limits available functions based on user privileges. For example, product image registration can be limited to users with administrator privileges, while general staff can only view the images. Data encryption is performed both at rest and in transit, using a strong encryption algorithm such as AES-256. Operation logs are also recorded, and a mechanism is implemented to detect and track unauthorized access and operations.

[0019] In one implementation, this system provides integration functionality with existing POS systems. By linking with the API of a standard POS system, it can synchronize product master data, link inventory information, and integrate sales data. It supports both real-time and batch synchronization as an integration method, allowing users to select the method based on their system requirements. It also takes into account integration with peripheral devices such as barcode scanners, receipt printers, and cash drawers, making it a comprehensive POS solution. Its data format conversion function also makes it easy to transfer data between different POS systems.

[0020] One implementation includes a performance optimization function. To speed up image processing, GPU acceleration can be utilized, enabling the use of frameworks such as CUDA, OpenCL, and Metal. Furthermore, a caching function for processed images reduces processing time by avoiding reprocessing the same product. Image compression technologies include JPEG, PNG, WebP, and other formats appropriately selected according to the product's characteristics, optimizing the balance between image quality and capacity. For network communication, efficient protocols such as HTTP / 2 and gRPC are used to optimize data transfer.

[0021] In one implementation, the user interface will employ a responsive design to accommodate a variety of device sizes. While primarily intended for use on tablet devices, the system will be designed to display and operate appropriately on different screen sizes, including desktop PCs, smartphones, and large displays. While providing an interface optimized for touch operation, it will also support operation with a mouse and keyboard. Accessibility features include text size adjustment, high contrast mode, and screen reader support, making the system easy to use for a variety of users.

[0022] One implementation provides an analysis and reporting function. Data such as the number of times a product is displayed, click rates, and correlation with sales is collected and analyzed, and used to optimize the most effective layout. The heat map function visualizes areas where customers' gazes are concentrated, helping to improve product placement. The implementation of an A / B testing function makes it possible to quantitatively compare the effectiveness of different layout patterns. Analysis results are visually displayed using graphs and charts, and can be exported in PDF or Excel format.

[0023] One implementation allows integration with augmented reality (AR) functionality. It provides a function to display products in an AR space using cropped product images. This enables new customer experiences, such as simulating product placement in a physical store or previewing products in online shopping. AR frameworks such as ARCore, ARKit, and WebXR are used, with the appropriate choice depending on the platform. It also includes a 3D model conversion function, making it possible to generate simple 3D representations from 2D images.

[0024] One implementation will include a continuous improvement function for the machine learning model. Data accumulated during system use will be used to retrain the image recognition model, improving recognition accuracy. Manually corrected cropping results from users will be used as feedback data to continuously improve the model's accuracy. By adopting federated learning technology, it will be possible to integrate data from multiple stores for learning while protecting privacy. Model updates will be implemented in a phased rollout manner to avoid impacting system operation.

[0025] As one embodiment, a voice interface function can be added. This allows product names and prices to be entered by voice, further improving work efficiency. Cloud services such as Google Cloud Speech-to-Text, Amazon Transcribe, and Microsoft Azure Speech Services can be used as the voice recognition engine, or a built-in voice recognition engine that operates offline can also be used. Layout operations via voice commands (such as "move this product to the right" and "enlarge the price") are also supported, allowing for hands-free operation.

[0026] As one of the implementation modes, the printing function will be enhanced. To print the created layout with high quality, functions such as conversion to CMYK color mode, optimization for print resolution, and bleed settings will be provided. By linking with printer drivers, output on various printers will be supported. In addition, a tiling function for large-format printing and output in catalog format spanning multiple pages will be supported. A print preview function will allow the actual print result to be checked in advance.

[0027] As one of the embodiments, a collaboration function is implemented. A real-time collaborative editing function is provided, allowing multiple users to edit the same layout at the same time. Low-latency synchronization is achieved using technologies such as WebSocket and WebRTC. An editing permission management function allows settings such as allowing only specific users to edit, while other users can only view. Comment and chat functions are integrated to facilitate communication when creating layouts. A visualization of change history makes it possible to track who made what changes and when.

[0028] As one embodiment, the system will expand its functionality for linking with external services. With the automatic posting function to social media, when a new product layout is created, it can be automatically posted to Facebook, Instagram, Twitter (now X), etc. By linking with an e-commerce site, it is also possible to reflect the created layout directly on the product page of the online store. Furthermore, by linking with a digital signage system, it can be automatically displayed on in-store displays. These linking functions are realized through the APIs of each service, and standard protocols such as OAuth 2.0 are used for authentication.

[0029] In one embodiment, the various technical elements used in this system are designed to be interchangeable. For example, even if TensorFlow is used as the image recognition engine, an abstraction layer is provided to allow easy switching to other frameworks such as PyTorch, ONNX Runtime, and Core ML. Similarly, the database system, cloud platform, and user interface framework are also designed to be independent of any specific technology, allowing the system to flexibly respond to technological advances and changing requirements.

[0030] In one embodiment, the step of photographing a product is performed by combining multiple photographing conditions and photographing techniques. In the preparation stage of the photographing environment, it is preferable to place a white or solid-color background board behind the product to create a uniform lighting environment, but this is not limited to this. Photographing under natural light or using the existing store environment as the background is also possible. A digital camera with a resolution of 12 megapixels or higher is recommended as the photographing device, but a wide range of photographing devices are compatible, from smartphone cameras with approximately 8 megapixels to high-resolution cameras with over 50 megapixels. When photographing, it is important to adjust the camera so that the entire product fits within the angle of view and to compose the image so that the product's outline is clearly discernible. The photographing function allows for the capture of images from multiple angles, such as front, side, and top views, and the optimal image can be selected or synthesized in subsequent processing.

[0031] In one embodiment, the processing step of using artificial intelligence to recognize product parts from captured images and separate them from the background is implemented as a multi-stage processing pipeline. In the first stage, image processing filters such as noise reduction, contrast adjustment, and edge enhancement are applied to the input image as preprocessing. These filters can use conventional image processing techniques such as Gaussian filters, median filters, and unsharp masks, but deep learning-based noise reduction networks can also be used. In the second stage, a semantic segmentation model is used to determine whether each pixel in the image belongs to the product or the background. The model can be selected from a variety of architectures, including U-Net, DeepLab v3+, Mask R-CNN, and SegFormer, and the appropriate model is selected based on the required processing speed and accuracy.

[0032] In one embodiment, the third stage of background separation involves post-processing the segmentation results. Initial segmentation results may contain artifacts such as edge imprecision, small holes, and isolated regions. To correct these artifacts, morphological operations (dilation, erosion, opening, and closing) are applied to improve the quality of the mask. Optimization techniques such as the GrabCut algorithm and Graph Cut can also be used to further improve edge accuracy. For transparent or translucent products, alpha matting techniques are applied to properly handle edge transparency. The processed results are saved in PNG or TIFF format with an alpha channel to preserve background transparency.

[0033] In one embodiment, the final processing stage involves quality evaluation and, if necessary, reprocessing. The automated quality evaluation system evaluates the cropped result by quantifying factors such as the smoothness of the edges, the completeness of the product area, and the presence or absence of remaining background elements. If the evaluation score falls below a threshold, the system automatically attempts reprocessing with different parameters or provides an interface that prompts the user to make manual corrections. The manual correction function allows users to make fine corrections using tools such as a brush, eraser, and border adjustment. A function is also implemented that displays multiple processing results in comparison, allowing the user to select the optimal result. All processing history is saved, allowing users to return to any stage and reprocess as needed.

[0034] In one embodiment, the multi-angle photography feature is implemented as an advanced photography system to comprehensively capture the product's three-dimensional shape information. The photography device can be configured in a variety of ways, including a fixed multi-camera array, a single-camera system with a rotating platform, or a handheld camera in continuous shooting mode. A multi-camera array uses a minimum of three to a maximum of 12 cameras positioned at different angles, capturing images simultaneously in all directions. A rotating platform system rotates the product in 15- to 30-degree increments, generating a set of 12 to 24 images. Camera parameters such as exposure, white balance, and focus are automatically optimized for each angle to ensure consistent image quality.

[0035] In one embodiment, the captured images from multiple angles are processed using advanced image analysis algorithms. First, each image is evaluated for quality, automatically detecting images with issues such as blur, out-of-focus, or poor exposure, and then either discarding them or prompting re-shooting. The quality assessment includes evaluation of sharpness using a Laplacian filter, exposure evaluation using histogram analysis, and contour clarity evaluation using edge detection. Next, a product feature matching algorithm (e.g., SIFT, SURF, ORB) is used to establish correspondences between each image and estimate the product's three-dimensional shape. This information plays an important role in selecting the optimal capture angle and merging multiple images.

[0036] In one embodiment, the optimal image selection process is automated using a machine learning-based evaluation model. The evaluation model assigns a score to each image from each angle, taking into account multiple criteria, such as product visibility, aesthetic quality, and information content. For example, frontal images emphasize the readability of product logos and labels, side images evaluate product thickness and shape characteristics, and top-view images evaluate overall proportions. Different evaluation criteria can also be applied to different product categories, allowing for flexible selection, such as prioritizing frontal labels for beverage products and oblique angles that enhance three-dimensionality for accessories. The evaluation results are visually displayed as a heat map, providing users with reference information for their final selection.

[0037] In one embodiment, multiple-image compositing combines various advanced image processing techniques. Image stitching technology seamlessly combines multiple overlapping images to create a 360-degree view of a product. Processing such as color adjustment, brightness correction, and geometric transformations are applied to achieve a natural-looking composite result. Focus stacking technology also generates a high-resolution image with in-focus focus from images captured at different focal lengths. HDR (High Dynamic Range) compositing combines images captured at different exposures to create an image with a rich range of tones from dark to bright. Furthermore, background blurring using depth information and super-resolution processing that integrates information from multiple angles can generate high-quality product images that cannot be achieved with a single image. These compositing processes are performed in real time or near real time using GPU acceleration, allowing users to instantly see the results and adjust parameters as needed.

[0038] One embodiment relates to a product layout free arrangement system using image recognition, which includes a photographing means for photographing products, an image processing means for recognizing product parts from the photographed image using artificial intelligence and separating them from the background, a display layout means for positioning the separated product images in any position, and a database means for associating and managing product information. The image processing means uses an image segmentation algorithm based on deep learning, and the display layout means includes a user interface for changing the position of product images using drag-and-drop operations. The system further includes a caching means for caching processed images, thereby achieving efficient processing. Conventional product management systems have limited performance in the photographing step, and the image processing means has insufficient accuracy in the step of recognizing product parts from the photographed image using artificial intelligence and separating them from the background. The step of generating a display layout for positioning separated product images in any position and the step of associating and managing product information are not coordinated, making it difficult to achieve integrated processing as a method for free product layout arrangement using image recognition. Even when implementing the function of acquiring photographed images of products on a computer, there were many programs that used image recognition to freely arrange products in a layout, but these programs did not fully integrate the functions of using artificial intelligence to recognize product parts from the acquired images and separate them from the background, generating a display layout that allows the separated product images to be positioned anywhere, and associating and managing product information. This is achieved by improving the performance of the photographing means for photographing products and by using an image segmentation algorithm based on deep learning in the image processing means that uses artificial intelligence to recognize product parts from the photographed image and separate them from the background. A system for free product layout placement using image recognition can be realized by incorporating a user interface that allows the position of product images to be changed by drag-and-drop operations into a display layout means that allows separated product images to be placed in any position, and further comprising a database means for associating and managing product information and a cache means for caching processed images. The system also improves the efficiency of a method for free product layout placement using image recognition, which includes steps from photographing products to recognizing product parts from the photographed image using artificial intelligence and separating them from the background, generating a display layout that allows separated product images to be placed in any position, and associating and managing product information, and optimizes a computer-implemented program for free product layout placement using image recognition.

[0039] This embodiment includes a multi-function camera unit integrating a high-resolution digital camera, LED ring lighting device, tripod mounting mechanism, and background control panel as the product image capture means. The image processing means, which uses artificial intelligence to recognize product features from captured images and separate them from the background, is based on a multi-layer convolutional neural network running on a GPU-accelerated processing platform. Its core is a four-stage processing system consisting of a pre-processing module, a feature extraction module, a segmentation execution module, and a post-processing module. The display layout means, which allows separated product images to be positioned anywhere, is implemented as a high-performance display processing system incorporating a hardware-accelerated rendering engine using WebGL technology, a real-time coordinate transformation system, a collision detection algorithm, and a snap function. The database means, which associates and manages product information, employs a hybrid configuration of a NoSQL database and a relational database, realizing a distributed storage architecture that efficiently manages product metadata, image binary data, layout information, and history data. The image processing means uses a deep learning-based image segmentation algorithm, specifically an improved semantic segmentation model based on the U-Net architecture. This model utilizes a ResNet-based pre-trained network as a feature extractor in the encoder, and performs high-precision reconstruction processing by combining attention and skip connections in the decoder. Multi-scale feature fusion technology improves product boundary detection accuracy at different resolutions, achieving accurate pixel-level product region extraction. The display layout means includes a user interface that allows product image placement to be changed by drag-and-drop operations, and is equipped with a multimodal input system that supports mouse, touch, and keyboard operations. Operation history management provides undo / redo, bulk movement, group selection, and placement guide display functions, creating an intuitive and efficient layout editing environment. The system also includes a cache means for caching processed images, integrating memory hierarchy optimization, prefetching, and compression / storage functions, significantly improving response time and memory utilization when handling large amounts of image data, achieving efficient processing. To improve the performance of the imaging system, we have built a next-generation imaging platform that integrates an automatic exposure control system, automatic color temperature correction function, focal length optimization algorithm, and depth-of-field control mechanism. This imaging system incorporates a machine learning-based image quality assessment model and optimizes imaging conditions in real time to ensure consistently high-quality product images. To achieve high accuracy, we developed a deep learning-based image segmentation algorithm for the image processing system, which uses artificial intelligence to recognize product features from captured images and separate them from the background. We developed a new hybrid model, CNN-ViT, that combines the Vision Transformer architecture with a convolutional neural network, achieving separation accuracy significantly superior to conventional methods. This model effectively utilizes contextual information from the entire image through a self-attention mechanism, simultaneously extracting local features and understanding global relationships, enabling accurate separation of products against complex backgrounds and with similar colors. The technical implementation includes a display layout method that allows users to freely position separated product images, as well as a user interface that allows users to change the position of product images using drag-and-drop operations. It integrates a physics engine for a natural user experience, real-time collision detection to prevent overlaps, a magnetic adsorption algorithm for alignment support, and a read-ahead processing function using an operation prediction system. These features enable intuitive and fast placement operations even when dealing with large volumes of product images, significantly improving user efficiency. To realize a free product layout system using image recognition, which also includes a database means for associating and managing product information and a cache means for caching processed images, we have built an innovative data management platform that combines distributed database technology, in-memory processing technology, and a blockchain-based consistency assurance mechanism. This system is capable of efficiently managing petabytes of image data and product information, and maintaining high responsiveness even when simultaneously accessed on a global scale. The process begins with photographing a product. In the next step, AI is used to recognize and separate product features from the background in the captured image. This step utilizes wavelet transform for pre-processing noise removal, and automatically performs image quality enhancement processing by combining an edge-preserving smoothing filter and gamma correction. In the image segmentation process, an uncertainty assessment function combining multiscale analysis and probabilistic inference is implemented to quantify the reliability of the separation results, efficiently achieving automatic quality control and, if necessary, manual verification. In the next step, the system generates a display layout that allows separated product images to be positioned anywhere. This function automatically suggests optimal layouts based on screen size using a responsive design algorithm, a layout evaluation system that takes aesthetic balance into account, and a visual optimization function based on color harmony theory. Finally, in the step of associating and managing product information, natural language processing is used to automatically extract product attributes, search for similar products, and automatically categorize the products, achieving fully automated metadata management. To improve the efficiency of the image recognition-based free product layout placement method, we adopted a parallel distributed processing architecture in a GPU cluster environment, improving the batch processing capacity of product images by more than 100 times compared to conventional methods. The processing load balancing system dynamically and optimally allocates processing resources using a machine learning-based load prediction algorithm, maintaining stable processing performance even during peak times. Furthermore, the processing result quality monitoring system combines statistical quality control techniques and anomaly detection algorithms to automatically monitor processing quality and provide early warning functions. To optimize the computer-implemented image recognition-based free product layout placement program, we adopted a microservices architecture to significantly improve the system's scalability and maintainability, and automatic scaling using container technology ensures flexible resource management in response to demand fluctuations. Furthermore, we built a continuous integration and continuous deployment (CI / CD) pipeline to enable rapid deployment of feature updates and bug fixes. This next-generation product management solution integrates and efficiently implements all functions, from capturing product images to using artificial intelligence to recognize and separate product parts from the background, generating display layouts that allow separated product images to be positioned anywhere, and associating and managing product information.

[0040] The image recognition-based product layout free arrangement system of this embodiment includes a multi-function photography unit integrating a high-resolution digital camera, an LED ring lighting device, a tripod fixing mechanism, and a background control panel as a photography means for photographing products. The image processing means, which uses artificial intelligence to recognize product parts from the photographed image and separate them from the background, has a core multi-layer convolutional neural network running on a GPU-accelerated processing platform, and is structured into a four-stage processing system consisting of a pre-processing module, a feature extraction module, a segmentation execution module, and a post-processing module. The display layout means has a dual-mode function that allows separated product images to be placed in any position or within a preset template. In the free placement mode, it is implemented as a high-performance display processing system incorporating a hardware-accelerated rendering engine using WebGL technology, a real-time coordinate conversion system, a collision detection algorithm, and a snap function, and includes a user interface that allows the placement of product images to be freely changed by drag-and-drop operations. In the template placement mode, it is equipped with a template library that allows users to select from preset layout patterns, providing a variety of placement patterns including category-specific templates, seasonal templates, purpose-specific templates, and industry-standard templates. The template function incorporates an optimal placement algorithm for each product category and automatically selects dedicated layout patterns that take product characteristics into account, such as templates for apparel, home appliances, food, and books. Each template integrates automatic product size adjustment, placement position optimization, and visual balance adjustment functions, improving the quality and efficiency of automatically placing product images in the template's specified position. Additionally, the template customization function allows users to create their own layout patterns and register them in the template library. The system's mode switching function allows instant switching between free layout mode and template layout mode with a single click, and when changing modes mid-work, editing can be continued in the new mode while retaining the existing layout information.When switching from free layout mode to template layout mode, the system is equipped with a function that automatically recommends the template most suitable for the current product layout, and when switching from template layout mode to free layout mode, free editing based on the template layout is possible. The database means for associating and managing product information employs a hybrid configuration of a NoSQL database and a relational database, realizing a distributed storage architecture that efficiently manages product metadata, image binary data, layout information, template information, and history data. Template usage history, placement pattern analysis data, and user operation logs are integrated and managed, providing placement pattern optimization and personalized recommendation functions using machine learning algorithms. The system also includes a caching means for caching processed images, integrating memory hierarchical optimization, prefetching, and compression and storage functions in both template application processing and arbitrary placement processing, significantly improving response performance and memory utilization efficiency when handling large amounts of image data, achieving efficient processing.

[0041] An AI image processing means operates by using artificial intelligence to recognize product parts from images taken by a means for photographing or uploading products. In a product layout free arrangement system using image recognition, which is equipped with a means for arranging separated product images and a database means for managing product information in association with each other, the image processing means includes a function for separating and cutting out product parts from the background. The system further includes a means for creating a composite photo from one or more product images. The arrangement means includes a function for arranging product images by applying them to a preset template, and a function for automatically generating a layout according to the characteristics of the products. The system is configured to include a user interface that allows product images to be freely arranged by drag-and-drop operations. The system further includes a caching unit for caching the processed image. A method for freely arranging products in a layout using image recognition is realized, which includes the steps of photographing or uploading products, recognizing product parts from the photographed image using artificial intelligence and processing the image with the AI, arranging the separated product images, and associating and managing product information. The image processing step includes a process of separating and cutting out product parts from the background and a process of creating a composite photo from multiple product images. A product layout free arrangement program using image recognition is constructed that realizes the following functions on a computer: acquiring or uploading photographed product images; recognizing product parts from the acquired images using artificial intelligence and performing image processing using AI; arranging the separated product images; and associating and managing product information. The system uses artificial intelligence to separate and cut out recognized product parts from the background of photographed or uploaded products, creating composite photos from single or multiple images. It integrates functions for arranging products by fitting them into preset templates, automatically generating layouts based on characteristics, and a user interface that allows free arrangement via drag-and-drop operations. A caching method for caching processed images enables an efficient image recognition-based free product layout system.

[0042] The means for photographing or uploading products operates a function for acquiring or uploading photographed images of products, and the means for recognizing product parts from the photographed image using artificial intelligence and performing image processing by AI executes a function for recognizing product parts from the acquired image using artificial intelligence and performing image processing by AI. The means for arranging separated product images provides a function for arranging separated product images, and the database means for associating and managing product information integrates a function for associating and managing product information, thereby realizing a program for free product layout arrangement using image recognition as a system for free product layout arrangement using image recognition. The image processing means includes a function for separating and cutting out product parts from the background, and a continuous processing flow is constructed from the step of photographing or uploading products to the step of recognizing product parts from the photographed image using artificial intelligence and performing image processing by AI. The system further includes a means for creating a composite photograph from one or more product images, and the means for creating the composite photograph functions as an integrated process that includes, in the image processing step, a process for separating and cutting out product portions from the background and a process for creating a composite photograph from multiple product images. In the step of arranging the separated product images, the arranging means executes a template arrangement process that includes a function for arranging the product images by fitting them into a preset template, and simultaneously operates an automatic layout generation process that includes a function for automatically generating a layout according to the product characteristics. The free arrangement function, which includes a user interface that allows product images to be freely arranged by drag-and-drop operations, provides a comprehensive solution as a method for free product layout arrangement using image recognition, while maintaining data consistency in the step of associating and managing product information. The system further comprises a caching means for caching processed images, thereby realizing efficient data access at all processing stages, from the image acquisition process in the step of photographing or uploading products, to the recognition process in the step of recognizing product parts from the photographed image using artificial intelligence and processing the image by AI, the arrangement process in the step of arranging separated product images, and the management process in the step of associating and managing product information. When the computer is made to realize the function of acquiring or uploading photographed product images, the function of arranging separated product images selects the optimal arrangement pattern in cooperation with the function of recognizing product parts from the acquired image using artificial intelligence and processing the image by AI, and the function of associating and managing product information operates as a product layout free arrangement program using image recognition that ensures metadata consistency. The system uses artificial intelligence to separate and crop recognized product images from the background of photographed or uploaded product images, and then uses single or multiple image data to create a single integrated composite photo from the separated images generated by this process. The system's layout function, which matches product images to pre-set templates, provides layout patterns optimized for each product category, while the system's automatic layout generation function, which performs dynamic layout optimization based on machine learning algorithms, allows product images to be freely positioned using drag-and-drop operations. The system's user interface allows for real-time position adjustment and visual feedback, and a caching method for caching processed images allows for both high-speed access and memory efficiency, creating an efficient image recognition-based free product layout system. The image acquisition process in the means for photographing or uploading products uses artificial intelligence to recognize product parts from captured images, and closely cooperates with preprocessing, feature extraction, and segmentation processes in the means for AI image processing, providing optimal input data for selecting an alignment algorithm in the means for arranging separated product images. A database means for associating and managing product information integrates and manages image processing results, alignment information, and metadata, and accumulates quality evaluation data and processing history when the image processing means executes the function of separating and cutting out product parts from the background. In the means for creating a composite photo from single or multiple product images, the system automatically performs color adjustment, size normalization, and alignment optimization processes for multiple images, and provides a function for dynamically selecting an output format corresponding to each of the template alignment, automatic generation, and free alignment functions of the arrangement means. It realizes a consistent workflow as a method for freely arranging products using image recognition, from verifying image quality in the step of photographing or uploading products, to high-precision recognition processing in the step of recognizing product parts from the photographed image using artificial intelligence and processing the image, providing various placement options in the step of arranging separated product images, and comprehensive data management in the step of associating and managing product information.The image processing step employs a high-precision separation algorithm that combines edge detection, region growing, and machine learning-based boundary recognition in the process of separating and cutting out product parts from the background, and is equipped with a compositing processing engine that integrates color space conversion, perspective conversion, and blending optimization in the process of creating a single composite photo from multiple product images. A program module that enables a computer to acquire or upload photographed product images works at the API level with a functional module that uses AI to recognize product parts from acquired images and processes the images using AI, and the functional module that arranges separated product images provides placement pattern analysis results that are reflected in the database access layer of a functional module that associates and manages product information, creating a program architecture for free product layout placement using image recognition. This program optimizes the image information of recognized and separated product parts from photographed or uploaded product data using AI technology through a processing routine that separates and cuts them out from the background and a processing routine that creates a composite photo from single or multiple image materials, and is implemented as a high-performance, highly scalable, and efficient free product layout placement system using image recognition, combining a user interface environment that integrates a pre-set template application function, an automatic layout generation function based on product characteristics, and a free placement function using drag and drop, with a cache management system that caches processed image data.

[0043] Examples and Modifications The components of the present invention are not limited to those described in the above embodiments, and can be combined, modified, substituted, or deleted in any manner within the scope of the technical concept of the present invention. The components described in each embodiment can be combined as appropriate as long as there is no technical contradiction, and the combination patterns are virtually infinite. Furthermore, each means, step, and functional unit of the present invention can be implemented independently or in any combination, and the effects obtained by such selective combinations are diverse. Furthermore, each parameter, threshold, and setting value can be changed as appropriate depending on the application and environment, and the range of possible changes and combinations are virtually limitless. The technical scope of the present invention is not limited to the individual embodiments described above, but also encompasses any combination, modification, or application of the components of each embodiment. The possibilities for such combinations are theoretically infinite, and those skilled in the art will be able to easily conceive of various modifications based on the technical concept of the present invention.

[0044] At the POS registers of restaurants, cafes, and fast food restaurants, food images are photographed or uploaded, and the food portion is recognized using AI, separated from the background, and the product image is displayed on the order screen. The food image is placed in a pre-set menu template, and price and calorie information is displayed synchronously using a database that associates and manages product information. At POS registers in supermarkets and convenience stores, the system uses artificial intelligence to recognize product parts from product images taken when scanning product barcodes, and displays the product photos on the register screen. A cache method that caches processed images enables high-speed display, making product confirmation more efficient for store clerks and customers. When a product tag is scanned at a clothing store's POS register, the color, size, and design of the clothing are visually displayed by arranging a separated product image.A user interface that allows product images to be freely arranged by drag and drop operation realizes a coordination suggestion function. At the POS register for home appliances, a single composite photo is created from multiple product images to display the product, accessories, and warranty all at once. The layout is automatically generated based on the product's characteristics, automatically displaying related products and recommended accessories to promote additional sales. At POS registers in dispensing pharmacies, prescription drugs and over-the-counter drugs are distinguished by photographing the drug package and cutting it out from the background. A database that associates and manages product information displays dosage, instructions, and precautions along with the image, supporting medication instruction. At the point-of-sale register for book sales, AI is used to recognize the product part from the book cover image and display it in conjunction with author information, price, and stock status. Book images are placed in pre-set templates, and a function is realized to automatically suggest related books and recommended products. At the POS registers of beauty salons and barber shops, photos are taken before and after treatment, and comparison photos are created from a single or multiple images. The processed images are cached and linked to the customer's medical record, allowing for efficient confirmation of treatment history and suggested services the next time the customer visits. At the POS register for pet supply stores, product images and pet size information are linked to visually display suitability. The system automatically generates layouts according to product characteristics and provides a function to automatically display recommended products by pet type and age. At a florist's POS register, images of bouquets and potted plants are captured, separated from the background, and displayed along with the price, flower language, and care instructions. By using drag-and-drop operations, arrangement proposals based on customer requests can be visually created. At the POS registers of sporting goods stores, AI is used to recognize product parts from images of equipment, and guidance on how to use it and selecting the right size is displayed with images. Related products and maintenance supplies are automatically suggested, encouraging bulk purchases. By utilizing these POS register systems, various industries will be able to realize the linked display of product images and product information, automatic suggestions for related products, and visual product confirmation functions, thereby creating a system for freely arranging product layouts using image recognition that can simultaneously improve sales efficiency and customer satisfaction.

[0045] The image recognition-based product layout free placement system of this embodiment is integrated into the POS register system of a restaurant, cafe, or fast food restaurant, and includes a high-resolution touchscreen camera module as a means for capturing or uploading food images. The AI ​​uses artificial intelligence to recognize food portions from the captured food images and process the images. The AI ​​employs a deep learning-based food recognition algorithm to accurately separate and crop the food portions from the plate and background. The separated product image placement function automatically places the food images in specified positions on the order screen and matches the food images to pre-set menu templates.

[0046] In the restaurant POS register system, a database means for managing product information in association with other information manages the name of the dish, price, calorie information, allergy information, and cooking time in an integrated manner, and automatically displays related information in synchronization with the image display. A caching means for caching processed images enables high-speed loading of menu images, shortening order processing time. A means for creating a composite photo from single or multiple dish images provides a function for displaying set menus and course meals all at once, and a function for automatically generating layouts according to the characteristics of the products enables effective placement of seasonal menus and limited-time menus.

[0047] In implementations in supermarket and convenience store POS register systems, a means for photographing or uploading a product is automatically activated when the product barcode is scanned, and an AI image processing means uses artificial intelligence to recognize the product parts from the photographed product image, and then separates the product packaging from the background. The image processing means extracts a front image of the product using a function to separate and cut out the product part from the background, and displays it in the product confirmation area on the register screen. A means for positioning the separated product image automatically selects a display position according to the product category, and displays it in conjunction with information such as price, product name, and production date.

[0048] In the retail store POS register system, a cache means for caching processed images prioritizes storing frequently sold product images in memory, shortening the processing time from barcode scanning to image display to less than 0.1 seconds. A database means for associating and managing product information integrates and manages JAN codes, product images, inventory information, and sales history, streamlining the product confirmation process for store clerks and customers. The system provides a visual display function for a list of purchased products by creating a composite photo from single or multiple product images when multiple products are scanned simultaneously.

[0049] In the implementation of a clothing store's POS register system, when the product tag is scanned, a means for photographing or uploading the product retrieves the corresponding image from a product image database, and a means for AI to recognize the product parts from the photographed image and automatically extract the color, size, and design features of the clothing. A means for arranging separated product images places the clothing image in the product information display area, visually displaying the stock status of products in different colors and sizes. A function including a user interface that allows product images to be freely arranged by drag-and-drop operations allows store staff to create a coordination suggestion screen according to customer requests.

[0050] In the apparel store POS register system, the function of matching product images to preset templates and arranging them provides patterns such as top and bottom sets, outerwear and innerwear combinations, and matching accessories.The function of automatically generating layouts according to product characteristics generates personalized product proposal layouts based on customer purchase history and preference information.A caching means for caching processed images efficiently manages images of seasonal and new products, and a database means for managing product information in association with each other displays information on brands, price ranges, and size ranges in an integrated manner.

[0051] In the implementation of a POS register system for home appliances, images of the home appliance itself, accessories, warranty, and instruction manual are acquired all at once by photographing or uploading the product, and the product parts are recognized from the photographed image using artificial intelligence, and each component is individually recognized by image processing by AI. A product confirmation screen that integrates the main body, accessories, and warranty is generated by creating a single composite photo from multiple product images, and the components are displayed in a logical arrangement pattern by arranging the separated product images.

[0052] In the electronics store POS register system, the function for automatically generating layouts according to product characteristics selects the optimal display layout according to the home appliance category (white goods, AV equipment, PC-related, etc.) and automatically displays related products and recommended accessories to promote additional sales. The function for arranging product images by applying them to preset templates visually suggests optional services such as extended warranties, installation services, and maintenance contracts. The database means for associating and managing product information integrates and manages technical specifications, prices, stock status, and customer reviews, and the cache means for caching processed images enables high-speed product information display.

Claims

1. A product layout free arrangement system using image recognition, which is integrated with a POS system and comprises: a means for photographing or uploading products; an image processing means for recognizing product parts from the photographed image using artificial intelligence and separating them from the background; a placement means for enabling the separated product image to be placed at any position on the product registration screen of the POS; and a database means for managing product information in association with product master data of the POS system.

2. The system for freely arranging product layouts using image recognition according to claim 1, characterized in that the image processing means includes a function for separating and cutting out product parts from the background, and the arrangement means includes a user interface that allows product images to be freely arranged by drag-and-drop operations.

3. The system for free product layout arrangement using image recognition according to claim 1 or 2, characterized in that the system further comprises a means for creating a composite photo from a single or multiple product images, and the composite photo can be displayed on the product registration screen.

4. 2. The system for free product layout arrangement using image recognition according to claim 1, wherein the arrangement means includes a function of arranging product images by fitting them to a template set in advance according to a product category.

5. The system for free product layout arrangement using image recognition according to claim 1, further comprising a caching means for caching processed images, thereby reducing processing time by avoiding reprocessing of the same product.

6. The system for free product layout placement using image recognition as described in claim 1, characterized in that the image processing means recognizes product parts using at least one of U-Net, Mask R-CNN, or DeepLab segmentation models.

7. A method for freely arranging products in a layout using image recognition, comprising the steps of: photographing or uploading a product; recognizing the product part from the photographed image using artificial intelligence and separating it from the background; placing the separated product image in a position of your choice on a product registration screen of a POS; and managing product information in association with product master data of the POS system.

8. The method for freely arranging product layouts using image recognition as described in claim 7, characterized in that the step of separating from the background includes a process of separating and cutting out the product portion from the background, and the step of arranging includes a process of freely arranging the product image by a drag-and-drop operation.

9. A program for free product layout arrangement using image recognition, which is characterized by realizing in a computer, as an integrated function with a POS system, a function of acquiring or uploading photographed images of products, a function of recognizing product parts from the acquired images using artificial intelligence and separating them from the background, a function of placing the separated product images in any position on the product registration screen of the POS, and a function of managing product information in association with product master data of the POS system.