Search

KR-20260064880-A - Template Auto-Matching System Recognition Method for Essential Data Management

KR20260064880AKR 20260064880 AKR20260064880 AKR 20260064880AKR-20260064880-A

Abstract

The present invention relates to a text recognition system that captures product packaging or image data using a mobile phone or similar device, automatically detects text areas without prior layout settings, and precisely extracts text. Based on the extracted text, it generates an automatic template and converts the data into a Key-Value format to detect whether essential items are missing. It performs complex tasks through a cloud-based processing module and processes and synchronizes data even offline. The purpose of this invention is to provide an automatic template matching system and a method for providing such a system that enables the automatic recognition and verification of nutritional information and raw material data indicated on product packaging, allows for cost reduction and increased user convenience by utilizing it with a standard mobile phone, and enables distributors to accurately manage data.

Inventors

  • 주현우

Assignees

  • 주식회사 딥파인

Dates

Publication Date
20260508
Application Date
20241030

Claims (9)

  1. In a template automatic matching system for essential data management, The above matching system is, An image capture module that allows a user to capture product packaging or image data using a camera on a standard mobile phone or other device; Text area detection module that detects text areas without prior layout settings; Text recognition extraction module (OCR) that extracts text detected in the above text area and precisely processes text of various sizes and fonts; A template generation module that automatically generates a template based on the extracted text above; and A Key-Value matching module that converts the text classified by the above template generation module into a Key-Value format, processes data regardless of the order of the keys, and detects errors if a set required item is missing; A template automatic matching system for essential data management characterized by including
  2. In claim 1, The above matching system is, A cloud-based processing module that compresses images captured from a camera of a standard mobile phone or other device and transmits them in real-time to a cloud server, performs complex text recognition and analysis tasks on the cloud server, and then transmits the processed data back to the standard mobile phone or other device; A notification module that automatically detects when the above-mentioned required items are missing and provides a warning to the user; and An offline processing module that temporarily processes data on the device regardless of network connection status and synchronizes data with the cloud when an internet connection is established; A template automatic matching system for essential data management characterized by further including
  3. In claim 1, The above Key-Value matching module is, Auto-labeling module; Includes more, The above auto-labeling module is, A template automatic matching system for essential data management characterized by automatically labeling items and classifying them into a set key through the text area detection module, the template generation module, and the key-value matching module, and automatically matching them in a key-value format.
  4. In claim 1, The image capture module is, An automatic template matching system for essential data management characterized by ensuring accurate data extraction by providing functions including focus adjustment, resolution correction, and preprocessing for images captured by a user.
  5. In claim 1, The above text area detection module is, Using deep learning-based Trace, Craft, and TrOCR models to detect text boundaries and separate the text-containing portion into bounding boxes, but including An image cropping module that removes unnecessary background from a detected text area, leaving only the necessary area; Includes more, An automatic template matching system for essential data management characterized by recognizing complex layouts and various text sizes and fonts, and detecting text areas to extract data.
  6. In claim 1, The above template generation module is, Using OpenAI LLM function calling, automatically extract the aforementioned required items by analyzing the location and layout of detected text without a pre-configured template, and automatically generate a template by analyzing new image data, but A user-customized template editing module that allows the user to modify the generated template within a set range; A template automatic matching system for essential data management characterized by further including
  7. In claim 3, The above Key-Value matching module is, The user sets the above-mentioned required item as the above-mentioned key, and the user changes the order or data format between key-value items, including, An automatic template matching system for essential data management characterized by being able to support multiple languages and respond to country-specific nutritional labeling requirements by linking with a cloud server.
  8. In claim 2, The above-mentioned cloud-based processing module is, It includes features such as real-time updating of various data-related regulatory changes and data-related matters, but A template automatic matching system for essential data management characterized by providing lightweight data transmission and cloud-based processing tailored to the processing capabilities of the device, processing data in offline mode regardless of network connection status, and synchronizing in real-time the data processed in the offline module of the camera of a standard mobile phone or other device when an online connection is established.
  9. In a method of providing by an automatic template matching system for essential data management, A step of capturing image data from product packaging using a camera of a standard mobile phone or other device; A step of detecting a text area without prior layout setting, removing an unnecessary background from the detected text area, and performing image cropping; A step of extracting text from the detected text area and performing text recognition extraction that precisely processes various sizes and fonts; A step of transmitting lightweight data to a cloud server and performing complex text recognition and analysis tasks on the cloud server; A step of setting and verifying required items configured through Key-Value matching; and A step of detecting data errors in real time and providing the error to the user as a notification; A method for providing an automatic template matching system for essential data management, characterized by including

Description

Template Auto-Matching System and Recognition Method for Essential Data Management The present invention relates to a template automatic matching system that automatically recognizes complex data, such as various information and raw materials indicated on product images, packaging, etc., and supports the user in systematically managing the said data. It automatically recognizes data of various layouts without prior setting and detects whether there are omissions or errors in essential data items. In addition, by applying a cloud-based lightweight deep learning model, it can be easily implemented even on regular mobile phones, providing a function that allows users to automatically match nutritional information on product packaging and look up necessary information in real time without separate complex equipment or systems. Today, food distributors face an increasing need to efficiently manage nutritional information and ingredient data for various products. Many countries mandate that nutritional information and ingredients be labeled on food packaging, with the aim of consumer protection by providing consumers with details such as allergens, nutritional content, and country of origin. Consequently, there is a growing need for distributors to provide consumers with accurate nutritional information while complying with legal requirements. Existing OCR and data processing systems rely on pre-specified image layouts and templates to operate, which limits their ability to recognize diverse layouts or complex product packaging. Large retailers handle a wide variety of products and brands, making it difficult to process unstructured data or changing layouts. They utilize a method of extracting text from each template area and matching it in a key-value format; however, accurate recognition is difficult without pre-setting templates, making additional user input essential for images containing complex layouts and large amounts of text, and data extraction is only possible if the user directly specifies the template area. Furthermore, existing systems that rely on expensive equipment or complex software are burdensome for small business owners and SMEs to adopt, and they also restrict user accessibility due to the need for specialized hardware. In such systems, operational inefficiency is significant and the likelihood of errors increases because data must be managed manually or complex equipment must be used. Providing accurate and reliable nutritional information is becoming increasingly important in terms of securing consumer trust. As consumers' interest in health and nutritional content grows, they want to know exactly what ingredients are in the food they purchase, and retailers must be able to provide accurate information to meet this demand. Finally, as global markets expand, distributors must be able to handle the diverse nutritional labeling requirements and linguistic differences of each country. Products distributed in the global market have different regulations and labeling formats depending on the nation, which increases the likelihood of errors during the data entry and verification processes to comply with national regulations; furthermore, handling these tasks manually requires a significant amount of time and manpower. Furthermore, different templates must be manually managed or configured for each country, and systems lacking multilingual support are prone to errors in nutritional information or ingredient labeling, and may fail to provide accurately translated data tailored to the specific requirements of each country. In addition, it is difficult to quickly update data in response to regulatory changes, and responses to such changes may be delayed. [Prior Art Literature] [Patent Literature] Republic of Korea Published Patent No. 10-2024-0034059 FIG. 1 is a diagram relating to the configuration of an entire system according to one embodiment of the present invention. FIG. 2 is a drawing for explaining the configuration of a device according to one embodiment of the present invention. FIG. 3 is a drawing for explaining a cloud server according to an embodiment of the present invention. FIG. 4 is a flowchart relating to a method for providing an automatic template matching system for essential data management according to an embodiment of the present invention. FIG. 5 is a configuration diagram of a text area detection module according to an embodiment of the present invention. FIG. 6 is a configuration diagram of a Key-Value matching module according to an embodiment of the present invention. FIG. 7 is a configuration diagram of a template generation module according to an embodiment of the present invention. FIG. 8 is an image of the result of detecting a text area in an image crop module according to an embodiment of the present invention. FIG. 8 is an image of code in which a function in a template generation module according to an embodiment of the present invention generates a prompt includ