General medical receipt data accurate identification method
Through the combination of computer identification module and OCR technology, the problems of low efficiency and poor accuracy caused by relying on manual identification of medical ticket data are solved, and efficient and accurate automated identification is achieved, reducing costs and enhancing the universality of the system.
Patent Information
- Application Number
- CN202411054269.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-19
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing medical receipt data identification relies on manual methods, resulting in low efficiency, poor accuracy, high cost, and easy omissions and errors.
The computer recognition module is used to identify ticket data, and text recognition is performed through image preprocessing and OCR technology combined with the recognition library, which evaluates the clarity of tickets and performs automated recognition, supporting the identification and processing of new templates.
It improves the accuracy and efficiency of medical receipt identification, reduces manual intervention and errors, reduces enterprise data processing and audit costs, and enhances the universality and convenience of the system.
Smart Images

Figure CN118982836A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of image and text recognition, and specifically is a method for accurately recognizing general medical receipt data. Background Art
[0002] Medical receipts usually refer to the fee receipts issued by medical institutions after providing medical services to patients. They include paper invoices and electronic invoices. They can be used as vouchers for medical expense reimbursement, the basis for verifying medical expenses, and legal evidence for any medical disputes or questions. The existing medical receipt data recognition and processing mainly rely on traditional manual methods. Manual recognition and processing of medical receipt data requires a lot of manpower and time. Omissions and errors are prone to occur during the manual recognition process, especially when dealing with large and complex receipt data. Long-term reliance on manual data recognition and processing not only increases the company's labor costs, but may also increase additional costs due to other problems caused by errors and delays. In this regard, we propose a universal medical receipt data accurate recognition method. Summary of the invention
[0003] In view of the deficiencies in the prior art, the present invention provides a method for accurately identifying universal medical receipt data to solve the above technical problems.
[0004] To achieve the above purpose, the present invention provides the following technical solution: a method for accurately identifying general medical receipt data, the method for identifying receipt data is:
[0005] S1. Input the information data needed from the receipt into the computer and put the receipt into the recognition module;
[0006] S2, the recognition module moves a single receipt to the recognition area for preliminary recognition, and evaluates the clarity of the receipt according to the set threshold. If the clarity of the receipt does not meet the requirements, the recognition is abandoned and the receipt is directly ejected from the ticket entrance. If the clarity of the receipt meets the requirements, image preprocessing is performed;
[0007] S3. After preprocessing, the recognition module compares the image with the receipt template in the recognition library. If the receipt is a template in the recognition library, text recognition is performed. If the receipt is not in the recognition library template, the text in the receipt image is divided into blocks, and a recognition template is made and retained in the recognition library, and then text recognition is performed;
[0008] S4. According to the input information data requirements, the recognized text information is searched to check whether the receipt image contains the required information data. If the receipt image contains the required information data, the information is extracted. After the extraction is completed, the recognition module pushes the receipt out of the ticket outlet. If the receipt image does not contain the required information data, the receipt is withdrawn from the ticket inlet and other information in the receipt is retained in the computer;
[0009] S5. After the receipt recognition is completed, the user enters the required information data entries and the table is automatically exported in the computer.
[0010] Preferably, the method for evaluating the clarity of the receipt includes:
[0011] Use edge detection function to identify the edge of the image, detect whether there is cropping or occlusion at the border of the receipt image, and determine the integrity of the receipt;
[0012] Use edge detection algorithms to extract text edges in images, analyze sharpness, and determine the clarity of text on receipts;
[0013] Combine pixel distribution and blur measurement methods to determine the overall clarity of the receipt image.
[0014] Preferably, the receipt image evaluation criteria are:
[0015] A1. Receipt completeness
[0016] Completely intact: 3 points;
[0017] There is cropping or occlusion, but the key information is complete: 2 points;
[0018] Part of the key information is missing, but does not affect the main identification: 1 point;
[0019] Key information is seriously missing and cannot be identified: 0 points;
[0020] A2. Text clarity
[0021] The text is clear, without blur or broken strokes: 3 points;
[0022] Some broken strokes are slightly blurred but do not affect recognition: 2 points;
[0023] The text edges are blurred, image processing is required, and recognition is difficult: 1 point;
[0024] The text is unclear and cannot be recognized: 0 points;
[0025] A3. Image clarity
[0026] The image is clear and noise-free: 3 points;
[0027] The image has slight noise or blur, which does not affect recognition: 2 points;
[0028] The image has a lot of noise or is obviously blurred, requires image processing, and is difficult to identify: 1 point;
[0029] The image is blurry and unrecognizable: 0 points.
[0030] Preferably, a comprehensive score of the clarity of the receipt is calculated based on the evaluation criteria of completeness, text clarity and image clarity;
[0031] Among them, the calculation formula for the comprehensive score of the receipt clarity is:
[0032] TS=x 1 × 1 +x 2 × 2 +x 3 × 3
[0033] Where TS represents the comprehensive score of the ticket clarity, x 1 、x 2 、x 3 They represent the receipt completeness score, text clarity score and image clarity score respectively, and w 1 、w 2 、w 3 Respectively represent the weights of receipt completeness, text clarity and image clarity.
[0034] Preferably, the detailed steps of comparing the receipt image with the recognition library in step S3 are:
[0035] Collect a large number of medical receipts, scan and retain the image text information, divide the data information area of the receipts by manual annotation, save them as recognition templates and build a recognition library;
[0036] The preprocessing of the receipt image that has passed the clarity assessment includes graying, binarization, and denoising. The feature matching algorithm is used to compare the preprocessed receipt image features with the template features in the recognition library. The similarity between the two is determined by the set threshold. If the similarity exceeds the threshold, the receipt is judged to be a template in the recognition library. If the similarity is lower than the threshold, the receipt is judged to be a new template.
[0037] When the receipt is a template in the recognition library, the OCR technology is directly used to perform text recognition in the corresponding area according to the input requirements;
[0038] If the receipt is not a template in the recognition library, perform layout analysis on the receipt image, identify text blocks, divide each text block into independent recognition units, and mark their positions and attributes. Save the image composed of the entire block information as a new template and add it to the recognition library. Then use OCR technology to perform text recognition in the corresponding area.
[0039] Preferably, the method for dividing the receipt image information area of the new template is:
[0040] Use edge detection algorithm to detect text edges in the image, use findcontours function to find contours in the image, and filter out text block contours based on the contour area, shape, direction and text edge;
[0041] For each text block, it is divided into independent recognition units according to the internal edges or contours, and the position information of each recognition unit is recorded, including the coordinates and bounding box in the image, and the attributes of the recognition unit are marked, including size and direction.
[0042] Preferably, in step S4, the identification information is firstly recognized in the area corresponding to the template in the recognition library. If the input text information cannot be found, the whole block recognition search is performed to determine whether the receipt image contains the required information data.
[0043] Preferably, during the recognition process, the recognition module automatically records the recognition status, timestamp and recognition result of each receipt, and stores them in the database for subsequent query and audit.
[0044] Preferably, the recognition module is connected to a computer to provide a manual comparison and verification page, in which the user verifies the questionable recognition data with the retained receipt image. If erroneous data is found, it is manually modified and the recognition module optimizes the recognition accuracy based on the modified information. In the absence of a physical receipt, the user sends the receipt image to the recognition module via a computer to complete the recognition.
[0045] Preferably, the information data input in step S5 includes but is not limited to at least one of the drug name, quantity, unit price, total price, purchase date, and customer information, and the user can select or combine the input information data according to actual needs.
[0046] Compared with the prior art, the present invention has the following beneficial effects:
[0047] The present application realizes automated and intelligent medical receipt recognition through the recognition module. By comprehensively evaluating the clarity of the receipt before recognition, it ensures that only the receipts that meet the clarity requirements are subject to subsequent recognition processing, thereby improving the overall recognition accuracy. During the recognition process, OCR technology is used in combination with the templates in the recognition library for text recognition, which further improves the accuracy and efficiency of recognition. The method can handle different types of medical receipts. Whether it is a template that already exists in the recognition library or a new template, it can be recognized through the corresponding processing flow, enhancing the versatility of the system. The user only needs to input the information data to be obtained, and the system will automatically complete the processes of receipt recognition, information extraction, and table export, reducing the user's manual operation and improving ease of use. Through automated recognition and processing processes, human intervention and errors are reduced, the error rate caused by human factors is reduced, the processing efficiency and accuracy are improved, and the company's costs for data processing and manual review are indirectly reduced. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 This is a flow chart of the method for accurately identifying general pharmaceutical receipt data. DETAILED DESCRIPTION
[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0050] The present invention provides a technical solution: a method for accurately identifying general medical receipt data, the method for identifying receipt data is:
[0051] S1. Input the information data needed from the receipt into the computer, such as the drug name, specification, quantity, price, purchase date, and customer information. Then, the user puts the medical receipt to be identified flatly into the ticket inlet of the identification module and starts the module to prepare for the next stage of processing.
[0052] S2. The high-precision scanning device built into the recognition module automatically moves a single receipt to a special recognition area for preliminary recognition. Advanced image processing technologies, such as edge detection and grayscale conversion, are used to quickly evaluate the clarity of the receipt. The preset clarity threshold is used to determine whether the receipt is suitable for further recognition. If the clarity of the receipt does not meet the requirements, recognition is abandoned and the receipt is directly ejected from the ticket entrance. If the clarity of the receipt meets the requirements, image preprocessing is performed to optimize the image quality through denoising, contrast enhancement, and binarization, laying a good foundation for subsequent text recognition.
[0053] S3. After preprocessing, the recognition module first attempts to compare the image with the stored medical receipt template in the recognition library. The template is built based on a large number of samples and can cover common medical receipt formats. If the receipt successfully matches a template in the library, the recognition strategy corresponding to the template will be directly used for text recognition. If no matching template is found in the library, the custom template generation process will be automatically entered, and the text in the receipt image will be divided into blocks, made into recognition templates and retained in the recognition library, and then text recognition will be performed;
[0054] S4. According to the input information data requirements, the recognized text information is retrieved. This process is realized through text search, which can quickly locate and confirm whether the receipt image contains the information data required by the user, and check whether the receipt image contains the required information data. If the receipt image contains the required information data, the information is extracted and sorted according to the format specified by the user. After the information extraction is completed, the recognition module pushes the receipt smoothly out of the ticket outlet through a mechanical device, and displays or sends the extracted information data to the user at the same time. If the receipt image does not contain the required information data, the receipt is withdrawn from the ticket inlet and other information in the receipt is retained in the computer;
[0055] S5. After receipt recognition is completed, the user enters the required information data entries, such as drug category and purchase time period, as screening conditions. Based on these conditions, all receipt information that meets the conditions is automatically summarized and a detailed table report is generated in the computer.
[0056] Further, the method for evaluating the clarity of the receipt includes:
[0057] Use edge detection function to identify the edge of the image, perform global scan on the image to identify edge information in the image. Edge information is not limited to the physical boundary of the receipt itself, but also includes the edges of text, lines, and patterns that may exist in the receipt. Through the detected edge information, the system analyzes whether the boundary of the receipt image is continuous and smooth, so as to determine whether there is cropping or occlusion at the boundary of the receipt image, and to determine the integrity of the receipt. If the boundary is discontinuous or obviously missing, the system determines that the integrity of the receipt is insufficient, which affects the subsequent recognition effect;
[0058] For the text area in the receipt, the edge detection algorithm is used to extract the text edge in the image, which can more accurately identify the subtle edges of the text strokes, providing a basis for subsequent analysis. By analyzing the sharpness, contrast and continuity characteristics of the extracted text edges, the system evaluates the clarity of the text on the receipt. If the text edge is fuzzy, the contrast is low or there is a break, the text clarity is determined to be insufficient.
[0059] The system performs statistical analysis on the pixels of the entire receipt image, including the distribution range of pixel values and standard deviation parameters. These parameters can reflect the overall brightness, contrast and detail richness of the image. The blur measurement method is used to calculate the blurriness of the image and determine the overall clarity of the receipt image. These methods evaluate the clarity of the image by comparing the similarity of adjacent pixels in the image or calculating the change in the image gradient. Images with high blurriness usually mean serious loss of details, which is not conducive to subsequent text recognition.
[0060] Furthermore, the receipt image evaluation criteria are:
[0061] A1. Receipt completeness
[0062] Completely intact: 3 points. The edges of the receipt are intact, without any cropping, damage or obstruction. All information areas are fully presented without any missing key information.
[0063] There is cropping or occlusion, but the key information is complete: 2 points. The edge of the receipt may be slightly irregular, or there may be slight cropping or occlusion, but the key information is still completely visible and does not affect the main recognition task;
[0064] Some key information is missing, but it does not affect the main identification: 1 point, some non-key information on the receipt is missing, but although the key information is incomplete, it is sufficient to support the main identification needs, and most of the transaction details can still be inferred through the remaining information;
[0065] Key information is seriously missing and cannot be recognized: 0 points. A large amount of key information on the receipt is missing or seriously damaged, making it impossible to extract valid transaction information from it, and the recognition task cannot be performed;
[0066] A2. Text clarity
[0067] Clear text, no blur or broken strokes: 3 points. The text on the receipt is clear and continuous, without blur or broken strokes. The text has a high contrast with the background and is easy to identify directly.
[0068] Slightly blurred or broken strokes, but does not affect recognition: 2 points. The text is clear as a whole, but some strokes are slightly blurred or broken. These defects can be improved through simple processing and do not affect the accuracy of the recognition results.
[0069] The text edges are fuzzy, image processing is required, and recognition is difficult: 1 point. The text edges are obviously fuzzy, and the boundaries between strokes are unclear. Complex image processing technology is required to improve the recognition rate. The recognition process is time-consuming and there are certain errors.
[0070] The text is blurred and unrecognizable: 0 points, the text is extremely blurred and almost unrecognizable, and even the most advanced image processing technology cannot accurately recognize it;
[0071] A3. Image clarity
[0072] The image is clear and noise-free: 3 points. The image quality of the receipt is high, the picture is clear, there are no noise or scratch interference factors, the color reproduction is accurate, and the details are rich.
[0073] The image has slight noise or blur, which does not affect recognition: 2 points. The image is relatively clear overall, but there are slight noise or blur. These defects do not affect the recognition of the main information, and there are only slight deficiencies in the processing of details.
[0074] The image has many noise points or is obviously blurred, which requires image processing and is difficult to identify: 1 point. The image has many noise points or obvious blurred areas, which requires image processing technology to improve clarity. The identification process is relatively complicated and involves certain risks.
[0075] The image is blurry and cannot be recognized: 0 points. The image quality is extremely poor, the picture is blurry, and it is almost impossible to recognize any valid information. The recognition task cannot be performed.
[0076] Furthermore, a comprehensive score of the receipt clarity is calculated based on the evaluation criteria of completeness, text clarity, and image clarity;
[0077] Among them, the calculation formula for the comprehensive score of the receipt clarity is:
[0078] TS=x 1 × 1 +x 2 × 2 +x 3 × 3
[0079] Where TS represents the comprehensive score of the ticket clarity, x 1 、x 2 、x 3 They represent the receipt completeness score, text clarity score and image clarity score respectively, and w 1 、w 2 、w 3 Respectively represent the weights of receipt completeness, text clarity and image clarity.
[0080] Furthermore, the detailed steps of the comparison process between the receipt image and the recognition library in step S3 are as follows:
[0081] Collect a wide range of medical receipt samples from multiple channels, retain the image text information, divide the data information area of the receipt by manual annotation, integrate the annotation information into the template based on the manual annotation results, generate a structured recognition template, and save the generated template to the recognition library for subsequent comparison;
[0082] The preprocessing of the receipt image that has passed the clarity assessment includes converting the color receipt image into a grayscale image to simplify the image data and improve the processing speed. In the grayscale process, weighted processing is usually performed according to the importance of the color channel to retain as much information as possible. The grayscale image is binarized to divide the pixels in the image into foreground and background. The selection of the binarization threshold needs to be adjusted according to the specific image characteristics to ensure the integrity and clarity of the text area. The binarized image is denoised using a filter or a denoising algorithm to eliminate noise and speckle interference factors in the image and improve the image quality. The preprocessed receipt image features are compared with the template features in the recognition library using a feature matching algorithm. The similarity between the two is determined by the set threshold. If the similarity exceeds the threshold, the receipt is judged to be a template in the recognition library. If the similarity is lower than the threshold, the receipt is judged to be a new template.
[0083] When the receipt is a template in the recognition library, the OCR technology is directly used to perform text recognition in the corresponding area according to the input requirements;
[0084] If the receipt is not a template in the recognition library, perform layout analysis on the receipt image, identify the text blocks, divide each text block into independent recognition units, mark their positions and attributes, save the image composed of the entire block information as a new template, add it to the recognition library, and then use OCR technology to perform text recognition in the corresponding area;
[0085] Furthermore, the method for dividing the receipt image information area of the new template is:
[0086] Use edge detection algorithm to detect text edges in images, and identify areas with significant brightness changes in images, i.e., text boundaries. The result of edge detection is a binary image, in which text edges are highlighted. Use findcontours function to find contours in images, and filter out text block contours based on contour area, shape, direction, and text edges. This function can identify all continuous pixel sets with similar brightness values in an image, i.e., contours. For receipt images, most of these contours correspond to text block boundaries.
[0087] For each text block, it is divided into independent recognition units according to the internal edges or contours. During the division process, it is necessary to ensure that each recognition unit is complete, that is, its bounding box can completely cover all the text in the unit, and record the position information of each recognition unit, including the coordinates and bounding box in the image, and mark the attributes of the recognition unit including size and direction. This information will be used in the subsequent OCR text recognition process to guide the recognition engine to correctly locate and recognize text.
[0088] Furthermore, in step S4, for the text recognition process of the receipt image, the system will first try to prioritize recognition in the area corresponding to the existing template in the recognition library. This is an efficient and accurate method because it directly uses known structured information to locate and extract key data. However, when the input text information cannot be found, the system needs to adopt a more comprehensive recognition strategy to ensure the integrity of the information, and change to a full-block recognition search to determine whether the receipt image contains the required information data.
[0089] Furthermore, during the recognition process, the recognition module automatically records the recognition status, timestamp and recognition results of each receipt to ensure that the receipt can be accurately tracked during the entire processing process. The accurate recording of timestamps not only helps to monitor the efficiency of the recognition process, but also provides valuable clues when problems arise, helping to quickly locate the cause of the problem and store it in the database for subsequent query and audit.
[0090] Furthermore, the recognition module is connected to a computer to provide a manual comparison and verification page, in which the user verifies the questionable recognition data with the retained receipt image. If erroneous data is found, it can be manually modified, and the recognition module optimizes the recognition accuracy based on the modified information. In the absence of a physical receipt, the user sends the receipt image to the recognition module via a computer to complete the recognition. This function greatly expands the application scenarios and convenience of the recognition module. The user only needs to use a scanner or mobile phone camera to take a picture of the receipt, and then send the image to the recognition system. After receiving the image, the recognition module will automatically start the recognition process and return the recognition result to the user. This method not only saves the user's time and energy, but also improves the efficiency and accuracy of the recognition process.
[0091] Furthermore, the information data input in step S5 includes but is not limited to at least one of the drug name, quantity, unit price, total price, purchase date, and customer information, and the user can select or combine the input information data according to actual needs. The user can select the information data of interest or need from the options provided by the system to view or export. The user can also combine different information data as needed to generate customized reports or analysis results.
[0092] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0093] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A method for accurately identifying general medical receipt data, characterized in that: The method of identifying receipt data is: S1. Input the information data needed from the receipt into the computer and put the receipt into the recognition module; S2, the recognition module moves a single receipt to the recognition area for preliminary recognition, and evaluates the clarity of the receipt according to the set threshold. If the clarity of the receipt does not meet the requirements, the recognition is abandoned and the receipt is directly ejected from the ticket entrance. If the clarity of the receipt meets the requirements, image preprocessing is performed; S3. After preprocessing, the recognition module compares the image with the receipt template in the recognition library. If the receipt is a template in the recognition library, text recognition is performed. If the receipt is not in the recognition library template, the text in the receipt image is divided into blocks, and a recognition template is made and retained in the recognition library, and then text recognition is performed; S4. According to the input information data requirements, the recognized text information is searched to check whether the receipt image contains the required information data. If the receipt image contains the required information data, the information is extracted. After the extraction is completed, the recognition module pushes the receipt out of the ticket outlet. If the receipt image does not contain the required information data, the receipt is withdrawn from the ticket inlet and other information in the receipt is retained in the computer; S5. After the receipt recognition is completed, the user enters the required information data entries and the table is automatically exported in the computer.
2. The method for accurately identifying general medical receipt data according to claim 1 is characterized in that: Methods for assessing the clarity of receipts include: Use edge detection function to identify the edge of the image, detect whether there is cropping or occlusion at the border of the receipt image, and determine the integrity of the receipt; Use edge detection algorithms to extract text edges in images, analyze sharpness, and determine the clarity of text on receipts; Combine pixel distribution and blur measurement methods to determine the overall clarity of the receipt image.
3. The method for accurately identifying general medical receipt data according to claim 2 is characterized in that: The evaluation criteria for receipt images are: A1. Receipt completeness Completely intact: 3 points; There is cropping or occlusion, but the key information is complete: 2 points; Part of the key information is missing, but does not affect the main identification: 1 point; Key information is seriously missing and cannot be identified: 0 points; A2. Text clarity The text is clear, without blur or broken strokes: 3 points; Some broken strokes are slightly blurred but do not affect recognition: 2 points; The text edges are blurred, image processing is required, and recognition is difficult: 1 point; The text is unclear and cannot be recognized: 0 points; A3. Image clarity The image is clear and noise-free: 3 points; The image has slight noise or blur, which does not affect recognition: 2 points; The image has a lot of noise or is obviously blurred, requires image processing, and is difficult to identify: 1 point; The image is blurry and unrecognizable: 0 points.
4. The method for accurately identifying general medical receipt data according to claim 3 is characterized by: Calculate the comprehensive score of the receipt clarity based on the evaluation criteria of completeness, text clarity and image clarity; Among them, the calculation formula for the comprehensive score of the receipt clarity is: TS=x1×w1+x2×w2+x3×w3 Where TS represents the comprehensive score of the receipt clarity, x1, x2, and x3 represent the receipt completeness score, text clarity score, and image clarity score, respectively; w1, w2, and w3 represent the weights of the receipt completeness, text clarity, and image clarity, respectively.
5. The method for accurately identifying general medical receipt data according to claim 1, characterized in that: The detailed steps of the comparison between the receipt image and the recognition library in step S3 are as follows: Collect a large number of medical receipts, scan and retain the image text information, divide the data information area of the receipts by manual annotation, save them as recognition templates and build a recognition library; The preprocessing of the receipt image that has passed the clarity assessment includes graying, binarization, and denoising. The feature matching algorithm is used to compare the preprocessed receipt image features with the template features in the recognition library. The similarity between the two is determined by the set threshold. If the similarity exceeds the threshold, the receipt is judged to be a template in the recognition library. If the similarity is lower than the threshold, the receipt is judged to be a new template. When the receipt is a template in the recognition library, the OCR technology is directly used to perform text recognition in the corresponding area according to the input requirements; If the receipt is not a template in the recognition library, perform layout analysis on the receipt image, identify text blocks, divide each text block into independent recognition units, and mark their positions and attributes. Save the image composed of the entire block information as a new template and add it to the recognition library. Then use OCR technology to perform text recognition in the corresponding area.
6. The method for accurately identifying general medical receipt data according to claim 5 is characterized in that: The method for dividing the receipt image information area of the new template is: Use edge detection algorithm to detect text edges in the image, use findcontours function to find contours in the image, and filter out text block contours based on the contour area, shape, direction and text edge; For each text block, it is divided into independent recognition units according to the internal edges or contours, and the position information of each recognition unit is recorded, including the coordinates and bounding box in the image, and the attributes of the recognition unit are marked, including size and direction.
7. The method for accurately identifying general medical receipt data according to claim 1, characterized in that: In step S4, the recognition information is first recognized in the corresponding area of the template in the recognition library. If the input text information cannot be found, the whole block recognition search is performed to determine whether the receipt image contains the required information data.
8. The method for accurately identifying general medical receipt data according to claim 1, characterized in that: During the recognition process, the recognition module automatically records the recognition status, timestamp and recognition results of each receipt, and stores them in the database for subsequent query and audit.
9. The method for accurately identifying general medical receipt data according to claim 1, characterized in that: The recognition module is connected to the computer and provides a manual comparison and verification page. The user verifies the questionable recognition data with the retained receipt image. If erroneous data is found, it can be manually modified. The recognition module optimizes the recognition accuracy based on the modified information. In the absence of the physical receipt, the user sends the receipt image to the recognition module via the computer to complete the recognition.
10. The method for accurately identifying general medical receipt data according to claim 1, characterized in that: The information data input in step S5 includes but is not limited to at least one of the drug name, quantity, unit price, total price, purchase date, and customer information, and the user can select or combine the input information data according to actual needs.
Citation Information
Patent Citations
Invoice information input method, device and equipment based on OCR technology and storage medium
CN109766778A
Bill processing method and device
CN111931473A
Batch OCR (optical character recognition) method and system for bills
CN115147855A
Image quality detection and image correction method for bills
CN116245883A
Shopping receipt identification method and system
CN116844167A
