A method and system for recognizing steel stamp characters

Through the stamp character recognition system combined with image capture and 3D scanning technology, the problem of difficult to identify blocked stamp characters in traditional technologies is solved, achieving higher recognition accuracy and efficiency, and providing strong support for drug packaging quality control.

CN118711188BActive Publication Date: 2024-11-12SHENZHEN RUIYIBO MEDICAL EQUIP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411189040.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-11-12
Estimated Expiration
2044-08-28

AI Technical Summary

Technical Problem

Traditional visual recognition technology is difficult to accurately identify the obscured steel-printed characters on the paper outer packaging of drugs, especially the coverage problems caused by the sealing of transparent tape, which affects the accuracy and reliability of the identification.

Method used

The image capture module is used to combine visual monitoring instruments and 3D scanners to construct orthoprojected images and three-dimensional point cloud data. Through image preprocessing, identification module, occlusion analysis module and comprehensive identification analysis module, the difficulty of stamped characters is identified and evaluated. Including reflection characteristic monitoring and reflection contrast calculation, the comparative analysis of the identification of the influence coefficient.

Benefits of technology

It improves the accuracy and efficiency of steel stamp characters recognition, reduces misidentification and omissions, provides strong technical support for drug packaging quality control and verification, and enhances the adaptability and accuracy of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118711188B_ABST
    Figure CN118711188B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for steel-printed character recognition, which relates to the field of image recognition technology. The occlusion analysis module can detect and identify the covering effect of transparent tape or other obstructions on steel-printed characters by monitoring reflection characteristics and calculating reflection contrast Fbdd. If the reflection contrast Fbdd is found to exceed a preset threshold, the system will issue an occlusion recognition instruction. This function effectively identifies and handles the recognition difficulties caused by occlusion. After receiving the occlusion recognition instruction, the comprehensive recognition analysis module calculates the recognition influence coefficient Syxs and performs a detailed analysis in combination with the three-dimensional point cloud data, which not only accurately evaluates the degree of overlap between the steel-printed character area and the occlusion area, but also helps identify possible influencing factors. The feedback module provides an objective judgment of the current recognition difficulty by comparing the evaluation threshold with the recognition influence coefficient Syxs. This comprehensive feedback mechanism can guide further adjustments and optimizations to improve the adaptability and accuracy of the system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a method and system for recognizing steel stamp characters. Background Art

[0002] In the field of information technology and automated control, image recognition systems are important tools for improving production efficiency and quality control. In particular, in the printing and packaging industries, image recognition technology is widely used for product identification and quality inspection. As a specialized image recognition application, steel stamp character recognition is dedicated to identifying and verifying steel stamp characters printed on the surface of objects. Steel stamp characters are often used to ensure product authenticity and tracking information, and are widely used on various documents, certificates, and packaging. In the pharmaceutical industry, steel stamp characters are often found on one side of the outer packaging of drugs and are often printed in an inwardly concave manner. They usually contain the production date, expiration date, and batch number of the drug.

[0003] Traditional visual recognition technology faces many challenges in recognizing the steel-printed characters on the paper outer packaging of medicines. Since the surface of the medicine packaging often has uneven steel-printed characters, and transparent tape is often used as a sealing tool for medicine boxes, it is often the case that the tape seals part of the steel-printed code, which makes it possible to cover the steel-printed characters when recognizing them. The above factors increase the complexity of character recognition. Traditional methods may have difficulty in accurately recognizing obscured steel-printed characters, which in turn affects the accuracy and reliability of recognition. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method and system for steel stamp character recognition, which solves the problems in the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a steel stamp character recognition system, including an image capture module, an image preprocessing module, a recognition module, an occlusion analysis module, a comprehensive recognition analysis module and a feedback module;

[0006] The image capture module is used to capture images of the outer packaging of the medicine using a visual monitoring instrument to obtain an orthographic projection image, and to monitor the surface of the outer packaging of the medicine in combination with a 3D scanner to construct a relevant three-dimensional point cloud data set;

[0007] The image preprocessing module is used to preprocess the orthographic projection image, remove noise, smooth the image, perform geometric correction, and adjust the image quality based on image enhancement technology;

[0008] The recognition module is used to identify the difference area based on the relevant three-dimensional point cloud data set, and set corresponding rectangular frames for several groups of difference areas, and identify the rectangular frame with the steel stamp characters according to the three-dimensional point cloud data in each group of rectangular frames;

[0009] The occlusion analysis module is used to divide the rectangular frame where the steel-printed characters are identified into several groups of image areas, and monitor the reflection characteristics to obtain the reflectivity Fsl in each image area, and construct the reflection contrast Fbdd between different image areas based on the reflectivity Fsl in each image area. If the reflection contrast Fbdd exceeds the threshold, a cover recognition instruction is issued externally;

[0010] The comprehensive recognition and analysis module is used to analyze the overlap factor Ccyz between the area in the rectangular frame where the steel stamp characters exist and the masked area after receiving the mask recognition instruction, and to construct the recognition influence coefficient Syxs in combination with the relevant three-dimensional point cloud data set;

[0011] The feedback module is used to pre-set an evaluation threshold value P, and compare and analyze the evaluation threshold value P with the recognition influence coefficient Syxs to estimate the difficulty of recognizing the steel-printed characters in the current drug outer packaging.

[0012] Preferably, the image capture module includes a capture unit and a three-dimensional state acquisition unit;

[0013] The capture unit is used to place the medicine on the workbench in advance, so that the surface of the medicine outer packaging is kept flat and fixed in position, and the outer packaging of the medicine is photographed by orthographic projection using a visual monitoring instrument to capture the orthographic projection image, wherein the visual monitoring instrument includes a GigE industrial camera and a network camera;

[0014] The three-dimensional state acquisition unit is used to use 3D scanning technology to scan the degree of depression on the surface of the outer packaging of the medicine, start the 3D scanner, emit a laser through the 3D scanner, and receive the reflected signal to obtain the three-dimensional point cloud data of the surface of the outer packaging of the medicine, use the point cloud registration technology, and splice and integrate the three-dimensional point cloud data of multiple scans to construct a related three-dimensional point cloud data set.

[0015] Preferably, the image preprocessing module includes a processing unit and an enhancement unit;

[0016] The processing unit is used to use the wavelet denoising technology to remove the noise component in the wavelet domain, and use histogram equalization to adjust the brightness, stretch the orthographic projection image range to adjust the contrast of the orthographic projection image; and compress the image data in PNG format;

[0017] The enhancement unit is used to achieve uniform grayscale distribution by using the image enhancement technology and CDF to remap the grayscale value of each pixel in the image.

[0018] Preferably, the recognition module includes a rectangular frame establishment unit and a steel stamp character area determination unit;

[0019] The rectangular frame establishment unit establishes a plane coordinate system based on the plane where the drug outer packaging is located, and identifies the depth area based on the relevant three-dimensional point cloud data set obtained by the 3D scanner, wherein the relevant three-dimensional point cloud data set refers to a set of points on the surface of the drug outer packaging, specifically including the position coordinates of each point on the drug outer packaging surface, the density of the points, the acquisition timestamp and the unique identifier of each point; at the same time, the distance from the 3D scanner to the plane coordinate system is set as a standard value. If the distance from the 3D scanner to the corresponding point on the surface of the drug outer packaging is greater than the standard value, it is preliminarily determined that the area corresponding to the corresponding point is an abnormal area, and corresponding rectangular frames are set for several groups of difference areas to obtain the number of difference areas Nys;

[0020] The stamp character area determination unit is used to determine the number of points greater than a standard value in each group of rectangular frames based on the three-dimensional point cloud data in each group of rectangular frames, and use the rectangular frame with the largest number of points as the rectangular frame where the stamp character exists.

[0021] Preferably, the occlusion analysis module includes a data testing unit, a reflection analysis unit, an early warning unit and a differentiation unit;

[0022] The data testing unit is used to divide the rectangular frame where the steel-printed characters are identified into a plurality of groups of image areas, and monitor the reflection characteristics of the plurality of groups of image areas using a spectroscopic reflectometer to obtain the reflectivity Fsl of all pixels in each image area;

[0023] The reflection analysis unit includes a local analysis subunit and an in-frame overall analysis subunit;

[0024] The local analysis subunit is used to obtain the mean reflectivity in each image area by combining a statistical averaging algorithm. The mean reflectivity in each image area Obtained by the following formula:

[0025]

[0026] Where i = 1, 2, 3, ..., n, n represents the total number of pixels in the image area, Fsl i Expressed as the reflectance of each pixel in the image area.

[0027] Preferably, the frame-wide overall analysis subunit is used to calculate the average reflectivity of each image area. Calculate the reflection contrast Fbdd between different areas, which can be obtained as follows:

[0028]

[0029] The significance of this formula is to reflect the area where the reflection characteristics of different regions vary, where: and are the mean reflectance values ​​of two different image regions;

[0030] The warning unit is used to pre-set a threshold value, and compares and analyzes the reflection contrast Fbdd between different areas with the threshold value to determine whether there is a reflection difference caused by obstacles in the rectangular frame where the steel-printed characters are stored. The specific comparison content is as follows:

[0031] If the reflection contrast Fbdd exceeds the threshold, it is determined that there is a reflection difference caused by obstacles in the rectangular frame containing the stamped characters, and a cover recognition instruction will be issued.

[0032] If the reflection contrast Fbdd does not exceed the threshold, it is determined that there is no reflection difference caused by obstacles in the rectangular frame where the stamped characters are stored;

[0033] The differentiation unit uses the Otsu algorithm to automatically determine the adaptation threshold, performs threshold processing on several groups of image areas, divides the image area into two parts, and takes the area where the reflectivity Fsl of each pixel is greater than the threshold as the foreground, and takes the area where the reflectivity Fsl of each pixel is less than the threshold as the background. Combined with the characteristics of the foreground, the area where the foreground is located is regarded as the obstacle occlusion area.

[0034] Preferably, the comprehensive identification and analysis module includes a regional cross-analysis unit and an impact unit;

[0035] The area intersection analysis unit is used to analyze the overlap factor Ccyz between the area in the rectangular frame where the steel stamp characters exist and the cover area after receiving the cover recognition instruction, which is specifically obtained according to the following formula:

[0036]

[0037] In the formula, A represents the area of ​​the blocked area, B represents the area of ​​the stamped character area, ∩ represents a logical "AND" operation, and ∪ represents a logical "OR" operation.

[0038] Preferably, the influence unit is used to establish a unit based on a rectangular frame, and then combine the overlap degree factor Ccyz obtained in the regional cross analysis unit, and after dimensionless processing, obtain the identification influence coefficient Syxs, which is specifically obtained according to the following formula:

[0039]

[0040] Where Nys represents the number of difference regions, Scc represents the depth difference, α and β are weight values, and C is the correction constant.

[0041] Preferably, the feedback module includes a comparison unit and an estimation unit;

[0042] The comparison unit is used to compare and analyze the recognition influence coefficient Syxs with the evaluation threshold P to estimate the difficulty of recognizing the steel stamp characters in the current drug outer packaging. The specific comparison content is as follows:

[0043] If the recognition influence coefficient Syxs ≥ the evaluation threshold P, it means that the recognition of the steel stamp characters in the current drug outer packaging is at an abnormal level, and the red identification light will be emitted three times in succession;

[0044] If the recognition influence coefficient Syxs is less than the evaluation threshold value P, it means that the recognition of the steel stamp characters in the current drug outer packaging is at a normal level, and a green identification light will be emitted once;

[0045] The evaluation unit is used to receive identification lights and take corresponding sorting operations according to identification lights of different colors. When a red identification light is received three times in a row, the corresponding drugs will be detained and classified into the area for quality inspection; when a green identification light is received once, the corresponding drugs will not be detained and the next inspection process will be continued.

[0046] A method for recognizing steel stamp characters comprises the following steps:

[0047] Step 1: Use visual monitoring equipment to capture images of the outer packaging of the drug to obtain an orthographic projection image, and use a 3D scanner to monitor the surface of the outer packaging of the drug to construct a relevant three-dimensional point cloud data set;

[0048] Step 2: preprocess the orthographic projection image to remove noise, smooth the image, perform geometric correction, and adjust the image quality based on image enhancement technology;

[0049] Step 3: Identify the difference areas based on the relevant 3D point cloud data set, set corresponding rectangular frames for several groups of difference areas, and identify the rectangular frames with the steel stamp characters based on the 3D point cloud data in each group of rectangular frames;

[0050] Step 4: Divide the rectangular frame where the stamped characters are identified into several groups of image areas, and monitor the reflection characteristics to obtain the reflectivity Fsl in each image area. According to the reflectivity Fsl in each image area, construct the reflection contrast Fbdd between different image areas. If the reflection contrast Fbdd exceeds the threshold, issue a cover recognition instruction.

[0051] Step 5: After receiving the cover recognition instruction, the overlap factor Ccyz between the area in the rectangular box where the steel stamp characters are identified in the recognition module and the cover area is analyzed, and the recognition influence coefficient Syxs is constructed in combination with the relevant three-dimensional point cloud data set;

[0052] Step 6: Pre-set the evaluation threshold P, and compare and analyze the evaluation threshold P with the recognition influence coefficient Syxs to estimate the difficulty of recognizing the steel stamp characters on the outer packaging of the current drug.

[0053] The present invention provides a method and system for recognizing steel stamp characters, which have the following beneficial effects:

[0054] (1) Through the image capture module combined with visual monitoring instruments and 3D scanners, the system can obtain the orthographic projection image and three-dimensional point cloud data of the outer packaging of the drug. This multi-dimensional data integration improves the comprehensive capture capability of complex surface structures, especially the capture of details of uneven steel-printed character surfaces, and enhances the basic data for recognition. The recognition module accurately identifies the difference areas based on the three-dimensional point cloud data set and divides these areas into rectangular frames, thereby accurately locating the areas where steel-printed characters exist. This method improves the recognition efficiency of steel-printed characters and reduces the possibility of misrecognition and omission. The occlusion analysis module can detect and identify the coverage effect of transparent tape or other obstructions on steel-printed characters through reflection characteristic monitoring and reflection contrast Fbdd calculation. If the reflection contrast Fbdd is found to exceed the preset threshold, the system will issue an occlusion recognition instruction. This function effectively identifies and handles the recognition difficulties caused by occlusion. After receiving the occlusion recognition instruction, the comprehensive recognition analysis module calculates the recognition influence coefficient Syxs and combines it with the three-dimensional point cloud data for detailed analysis. This not only accurately evaluates the overlap between the steel-printed character area and the occlusion area, but also helps identify possible influencing factors. The feedback module provides an objective judgment of the current recognition difficulty by comparing the evaluation threshold with the recognition influence coefficient Syxs. This comprehensive feedback mechanism can guide further adjustments and optimizations to improve the adaptability and accuracy of the system. In general, the system effectively improves the recognition accuracy and efficiency of steel-stamped characters on the outer packaging of medicines by comprehensively using technical means such as multi-dimensional data capture, precise preprocessing, accurate recognition, occlusion analysis and comprehensive feedback, providing strong technical support for the quality control and verification of medicine packaging.

[0055] (2) By setting the distance from the 3D scanner to the plane coordinate system as the standard value, the system can accurately determine the abnormal area whose distance is greater than the standard value. This method helps to distinguish between normal areas and abnormal areas on the outer packaging of medicines, thereby more accurately locating the area of ​​the stamped characters. The stamped character area determination unit calculates the number of points in each rectangular box that are greater than the standard value based on the three-dimensional point cloud data in each group of rectangular boxes. The system identifies the rectangular box with the most points as the area where the stamped characters exist. This method ensures that the identified area is highly consistent with the actual position of the stamped characters, further improving the accuracy and reliability of stamped character recognition.

[0056] (3) The reflection analysis unit uses a statistical mean algorithm through the local analysis sub-unit to calculate the mean reflectivity in each image area, so that the system can obtain the overall reflection characteristics of the image area, which is convenient for further analysis of whether there is a covering. The early warning unit uses the reflection characteristic differences detected by the analysis results. If it is found that the reflectivity does not meet the preset standard, the system can automatically issue a cover recognition instruction, which helps to timely discover and deal with occlusion problems, thereby improving the recognition rate of steel stamped characters. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a block diagram of a steel stamp character recognition system of the present invention;

[0058] Figure 2 The present invention is a schematic flow chart of a method for recognizing steel stamp characters. DETAILED DESCRIPTION

[0059] 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.

[0060] Example 1

[0061] See also Figure 1 , the present invention provides a steel stamp character recognition system, including an image capture module, an image preprocessing module, a recognition module, an occlusion analysis module, a comprehensive recognition analysis module and a feedback module;

[0062] The image capture module is used to capture images of the outer packaging of the medicine using a visual monitoring instrument to obtain an orthographic projection image, and to monitor the surface of the outer packaging of the medicine in combination with a 3D scanner to construct a relevant three-dimensional point cloud data set;

[0063] The image preprocessing module is used to preprocess the orthographic projection image, remove noise, smooth the image, perform geometric correction, and adjust the image quality based on image enhancement technology;

[0064] The recognition module is used to identify the difference area based on the relevant three-dimensional point cloud data set, and set corresponding rectangular frames for several groups of difference areas, and identify the rectangular frame with the steel stamp characters according to the three-dimensional point cloud data in each group of rectangular frames;

[0065] The occlusion analysis module is used to divide the rectangular frame where the steel-printed characters are identified into several groups of image areas, and monitor the reflection characteristics to obtain the reflectivity Fsl in each image area, and construct the reflection contrast Fbdd between different image areas based on the reflectivity Fsl in each image area. If the reflection contrast Fbdd exceeds the threshold, a cover recognition instruction is issued externally;

[0066] The comprehensive recognition and analysis module is used to analyze the overlap factor Ccyz between the area in the rectangular frame where the steel stamp characters exist and the masked area after receiving the mask recognition instruction, and to construct the recognition influence coefficient Syxs in combination with the relevant three-dimensional point cloud data set;

[0067] The feedback module is used to pre-set an evaluation threshold value P, and compare and analyze the evaluation threshold value P with the recognition influence coefficient Syxs to estimate the difficulty of recognizing the steel-printed characters in the current drug outer packaging.

[0068] During the operation of this system, the system uses visual monitoring instruments to obtain high-quality orthographic projection images, and combines 3D scanners to monitor the surface of the outer packaging of medicines to construct a detailed three-dimensional point cloud data set. This data fusion technology can capture the detailed information and surface features of the steel-printed characters, laying a solid foundation for subsequent character recognition and analysis. The image preprocessing module significantly improves the image quality by removing noise, smoothing images, geometric correction and image enhancement. The optimized image can show the steel-printed characters more clearly, which helps to accurately extract character features and reduce errors in the subsequent recognition process. The recognition module identifies the difference area based on the three-dimensional point cloud data set and sets the corresponding rectangular frame to locate the steel-printed characters. This process can accurately identify the area where the steel-printed characters exist, improving the accuracy and reliability of character recognition. The occlusion analysis module obtains the reflectivity of each area by monitoring the reflection characteristics of the image area, and calculates the reflection contrast Fbdd between different areas. If the reflection contrast Fbdd exceeds the set threshold, the system can issue a cover recognition instruction in time. This function helps to identify the area blocked by transparent tape or other objects, thereby reducing the impact of occlusion on character recognition. After receiving the cover recognition instruction, the comprehensive recognition analysis module analyzes the overlap factor between the steel-stamped character rectangle and the occluded area, and combines the 3D point cloud data to construct the recognition influence coefficient. This analysis can accurately evaluate the impact of occlusion on character recognition and provide a detailed recognition difficulty assessment. The feedback module compares and analyzes the set evaluation threshold with the recognition influence coefficient to estimate the difficulty of recognizing the steel-stamped characters in the current drug packaging. This mechanism provides a basis for dynamic adjustment and optimization, enabling the system to adapt to different recognition conditions and environmental changes.

[0069] Example 2

[0070] Please refer to Figure 1 ,Specifically: the image capture module includes a capture unit and a three-dimensional state acquisition unit;

[0071] The capture unit is used to place the medicine on the workbench in advance, so that the surface of the medicine outer packaging is kept flat and fixed in position, and the outer packaging of the medicine is photographed by orthographic projection using a visual monitoring instrument to capture the orthographic projection image, wherein the visual monitoring instrument includes a GigE industrial camera and a network camera;

[0072] The three-dimensional state acquisition unit is used to use 3D scanning technology to scan the degree of depression on the surface of the outer packaging of the medicine, start the 3D scanner, emit a laser through the 3D scanner, and receive the reflected signal to obtain the three-dimensional point cloud data of the surface of the outer packaging of the medicine, use the point cloud registration technology, and splice and integrate the three-dimensional point cloud data of multiple scans to construct a related three-dimensional point cloud data set.

[0073] The image preprocessing module includes a processing unit and an enhancement unit;

[0074] The processing unit is used to use the wavelet denoising technology to remove the noise component in the wavelet domain, and use histogram equalization to adjust the brightness, stretch the orthographic projection image range to adjust the contrast of the orthographic projection image; and compress the image data in the PNG format. The PNG format is conducive to compressing the image data without losing any information;

[0075] The enhancement unit is used to remap the grayscale value of each pixel in the image through image enhancement technology and using CDF to achieve uniform grayscale distribution. Cumulative Distribution Function (CDF) is a function used to describe the distribution of values ​​of a random variable. In image processing, CDF is used to describe the distribution of pixel grayscale values ​​in an image. Specifically, CDF represents the proportion of pixels in an image that are less than or equal to a certain grayscale value.

[0076] In this embodiment, by placing the medicine on a workbench so that the surface of its outer packaging is flat and fixed, visual monitoring instruments such as GigE industrial cameras and network cameras are used for orthographic projection shooting. This method ensures the stability and consistency of the image and helps to capture high-quality orthographic projection images of the outer packaging of the medicine. Advanced 3D scanning technology is used, laser scanning and point cloud registration technology are used to accurately scan the degree of depression on the surface of the outer packaging of the medicine. By integrating the three-dimensional point cloud data of multiple scans, the three-dimensional point cloud data set of the outer packaging of the medicine can be accurately constructed, providing rich depth information for subsequent character recognition. Wavelet denoising technology is used to effectively remove image noise and improve image quality. At the same time, histogram equalization is used to adjust the image brightness and contrast, stretch the grayscale range of the image, and make the steel-printed characters more prominent. In addition, the use of PNG format to compress image data can reduce the file size without losing information and retain the complete details of the image. By remapping the grayscale value of each pixel in the image through image enhancement technology and CDF, a uniform grayscale distribution is achieved. This processing makes the steel-printed characters in the image visually clearer and more prominent, further improving the accuracy of character recognition. In summary, the system uses advanced technologies and methods in the image capture and preprocessing stages to further improve the image quality and information integrity. Accurate image capture ensures the detailed orthographic projection and comprehensive acquisition of three-dimensional features of the drug packaging, while efficient image preprocessing provides a clear and noise-free image foundation for subsequent steel stamp character recognition. These optimization measures effectively improve the overall performance and accuracy of the steel stamp character recognition system.

[0077] Example 3

[0078] Please refer to Figure 1 ,Specifically: the recognition module includes a rectangular frame establishment unit and a steel stamp character area determination unit;

[0079] The rectangular frame establishment unit establishes a plane coordinate system based on the plane where the outer packaging of the drug is located, and identifies the depth area based on the relevant three-dimensional point cloud data set obtained by the 3D scanner, wherein the relevant three-dimensional point cloud data set refers to a set of points on the surface of the outer packaging of the drug, specifically including the position coordinates of each point on the surface of the outer packaging of the drug, the density of the points, the timestamp of the acquisition, and the unique identifier of each point. The density of the points refers to the number of points in a unit volume or unit area, which describes the density of the point cloud data; at the same time, the distance from the 3D scanner to the plane coordinate system is set as a standard value. If the distance from the 3D scanner to the corresponding point on the surface of the outer packaging of the drug is greater than the standard value, it is preliminarily determined that the area corresponding to the corresponding point is an abnormal area, and corresponding rectangular frames are set for several groups of difference areas to obtain the number of difference areas Nys;

[0080] The stamp character area determination unit is used to determine the number of points greater than a standard value in each group of rectangular frames based on the three-dimensional point cloud data in each group of rectangular frames, and use the rectangular frame with the largest number of points as the rectangular frame where the stamp character exists.

[0081] In this embodiment, the rectangular frame establishment unit establishes a plane coordinate system on the plane where the outer packaging of the drug is located, and identifies the depth area based on the three-dimensional point cloud data set obtained by the 3D scanner. By setting the distance from the 3D scanner to the plane coordinate system as the standard value, the system can further effectively identify the depth change of the outer packaging surface of the drug. If the distance of some points is greater than the standard value, the system automatically marks the area where these points are located as an abnormal area, and sets a rectangular frame for these difference areas. This method ensures the accurate identification of the depth abnormal area and lays the foundation for the subsequent steel stamp character analysis. The steel stamp character area determination unit counts the number of points greater than the standard value in each rectangular frame based on the three-dimensional point cloud data in the rectangular frame. The system determines the rectangular frame with the most points as the steel stamp character area by comparing the number of points in each group of rectangular frames. This process effectively screens out the area where the steel stamp characters exist, reduces the possibility of misidentification, and improves the accuracy of recognition. The system obtains the number of difference areas by setting multiple rectangular frames and performs detailed analysis in these areas. This method not only improves the ability to distinguish different areas, but also enhances the system's ability to recognize various details on the surface of the outer packaging of the drug, making the detection of steel stamp characters more comprehensive and accurate. In summary, the recognition module of the system further improves the recognition accuracy and efficiency of steel-printed characters by accurately establishing a rectangular frame and determining the steel-printed character area.

[0082] Example 4

[0083] Please refer to Figure 1,Specifically: the occlusion analysis module includes a data testing unit, a reflection analysis unit, an early warning unit and a differentiation unit;

[0084] The data testing unit is used to divide the rectangular frame where the steel-printed characters are identified into a plurality of groups of image areas, and monitor the reflection characteristics of the plurality of groups of image areas using a spectroscopic reflectometer to obtain the reflectivity Fsl of all pixels in each image area;

[0085] The reflection analysis unit includes a local analysis subunit and an in-frame overall analysis subunit;

[0086] The local analysis subunit is used to obtain the mean reflectivity in each image area by combining a statistical averaging algorithm. The mean reflectivity in each image area Obtained by the following formula:

[0087]

[0088] Where i = 1, 2, 3, ..., n, n represents the total number of pixels in the image area, Fsl i Expressed as the reflectance of each pixel in the image area.

[0089] In this embodiment, the data testing unit divides the identified steel stamp character rectangular frame into several groups of image areas and uses a spectroscopic reflectometer to monitor the reflection characteristics of these areas, so as to obtain the reflectivity Fsl of all pixels in each image area in detail. This detailed reflection characteristic monitoring provides rich data support, enabling the system to accurately analyze the reflection conditions of different areas. The local analysis subunit and the overall analysis subunit in the frame under the reflection analysis unit calculate the mean reflectivity in each image area by combining a statistical averaging algorithm. Through mean calculation, the system can clearly reflect the overall reflection characteristics of the image area, helping to distinguish different areas, such as the reflection difference between the steel-printed character area and the transparent tape-covered area. By comprehensively analyzing the reflection characteristics of the image area, the system can further accurately identify and distinguish the steel-printed character area and possible occlusion areas. This reflection characteristic analysis helps to determine whether there is occlusion, further reducing the omission of some steel-printed characters, thereby improving the system's recognition rate of steel-printed characters. In summary, through effective reflection characteristic monitoring and reflectivity mean analysis, the system can warn of possible occlusions, ensuring the efficiency and accuracy of the system in practical applications.

[0090] Example 5

[0091] Please refer to Figure 1 Specifically: the frame overall analysis subunit is used to calculate the average reflectivity of each image area. Calculate the reflection contrast Fbdd between different areas, which can be obtained as follows:

[0092]

[0093] The significance of this formula is to reflect the area where the reflection characteristics of different regions vary, where: and are the mean reflectance values ​​of two different image regions;

[0094] The warning unit is used to pre-set a threshold value, and compares and analyzes the reflection contrast Fbdd between different areas with the threshold value to determine whether there is a reflection difference caused by obstacles in the rectangular frame where the steel-printed characters are stored. The specific comparison content is as follows:

[0095] If the reflection contrast Fbdd exceeds the threshold, it is determined that there is a reflection difference caused by obstacles in the rectangular frame containing the stamped characters, and a cover recognition instruction will be issued.

[0096] If the reflection contrast Fbdd does not exceed the threshold, it is determined that there is no reflection difference caused by obstacles in the rectangular frame where the stamped characters are stored;

[0097] The differentiation unit uses the Otsu algorithm to automatically determine the adaptation threshold, performs threshold processing on several groups of image areas, divides the image area into two parts, the foreground and the background, and regards the area where the reflectivity Fsl of each pixel is greater than the threshold as the foreground, and the area where the reflectivity Fsl of each pixel is less than the threshold as the background. Combined with the characteristics of the foreground, the foreground area is regarded as the obstacle occlusion area, and the background area is regarded as the area where the steel-printed characters exist, because both the foreground area and the background area are within the rectangular frame where the steel-printed characters are identified. Among them, for the foreground in image processing, its characteristics usually refer to the area of ​​the target object in the image, and relative to the background, these areas are just the focus of the current processing.

[0098] Among them, the Otsu algorithm selects the best threshold based on the principle of maximizing the between-class variance. The algorithm divides the grayscale of the image so that the between-class variance between the foreground and the background is maximized, thereby obtaining the best segmentation effect; the Otsu algorithm is an automatic thresholding method, which is widely used in image segmentation in image processing. Its main purpose is to automatically select a relatively suitable threshold according to the grayscale histogram of the image to divide the image into foreground and background.

[0099] In this embodiment, the overall analysis subunit in the frame calculates the reflection contrast Fbdd between different areas based on the mean reflectivity in each image area, so as to further reflect the difference in reflection characteristics between different areas and help identify the reflection changes caused by obstructions such as transparent tape. The early warning unit compares the calculated reflection contrast with the preset threshold to determine whether there is a reflection difference caused by the obstruction of obstacles. If the reflection contrast Fbdd exceeds the threshold, the system issues a cover recognition instruction, indicating that there may be an obstruction phenomenon. If it does not exceed the threshold, it means that there is no significant obstruction. This intelligent early warning mechanism can effectively remind users to check the obstruction problem that may affect the recognition of steel-printed characters, thereby improving the accuracy of recognition. The differentiation unit uses the Otsu algorithm to automatically determine the adaptation threshold, threshold the image area, and divide the image into two parts: foreground and background. Specifically, the area with a reflectivity greater than the threshold is regarded as the foreground that may be an obstruction area, and the area with a reflectivity less than or equal to the threshold is regarded as the background that may be a steel-printed character area. This processing method ensures the effective separation of the foreground, i.e., the potential obstruction area, and the background, i.e., the steel-printed character area, and further enhances the accurate recognition of the obstruction. In summary, the system not only improves the accuracy of steel stamp character recognition through detailed reflection contrast calculation, intelligent early warning mechanism and automatic threshold processing, but also effectively identifies and handles potential occlusion problems. This system can provide more reliable drug packaging inspection and ensure the accuracy and effectiveness of the recognition results.

[0100] Example 6

[0101] Please refer to Figure 1 ,Specifically: the comprehensive identification and analysis module includes a regional cross analysis unit and an ,influence unit;

[0102] The area intersection analysis unit is used to analyze the overlap factor Ccyz between the area in the rectangular frame where the steel stamp characters exist and the cover area after receiving the cover recognition instruction, which is specifically obtained according to the following formula:

[0103]

[0104] In the formula, A represents the area of ​​the blocked area, B represents the area of ​​the stamped character area, ∩ represents a logical "AND" operation, and ∪ represents a logical "OR" operation.

[0105] The influence unit is used to establish a unit based on a rectangular frame, and then combined with the overlap degree factor Ccyz obtained in the regional cross analysis unit, and after dimensionless processing, the identification influence coefficient Syxs is obtained, which is specifically obtained according to the following formula:

[0106]

[0107] Wherein, Nys represents the number of difference regions, Scc represents the depth difference, α and β are weight values, 0<α≤1, 0<β≤1, and α+β=1, and C is the correction constant.

[0108] The depth difference Scc refers to the maximum and minimum distances from the 3D scanner to the corresponding points on the outer packaging surface of the drug, extracted from the rectangular frame where the steel stamp characters are identified and under the condition that the distance from the 3D scanner to the corresponding points on the outer packaging surface of the drug is greater than the standard value, and the maximum distance is subtracted from the minimum distance to obtain the depth difference Scc;

[0109] The feedback module includes a comparison unit and an estimation unit;

[0110] The comparison unit is used to compare and analyze the recognition influence coefficient Syxs with the evaluation threshold P to estimate the difficulty of recognizing the steel stamp characters in the current drug outer packaging. The specific comparison content is as follows:

[0111] If the recognition influence coefficient Syxs ≥ the evaluation threshold P, it means that the recognition of the steel stamp characters in the current drug outer packaging is at an abnormal level, and the red identification light will be emitted three times in succession;

[0112] If the recognition influence coefficient Syxs is less than the evaluation threshold value P, it means that the recognition of the steel stamp characters in the current drug outer packaging is at a normal level, and a green identification light will be emitted once;

[0113] The evaluation unit is used to receive identification lights and take corresponding sorting operations according to identification lights of different colors. When a red identification light is received three times in a row, the corresponding drugs will be detained and classified into the area for quality inspection; when a green identification light is received once, the corresponding drugs will not be detained and the next inspection process will be continued.

[0114] In this embodiment, after receiving the cover recognition instruction, the regional intersection analysis unit can accurately analyze the overlap factor Ccyz between the steel stamp character area and the occlusion area. This analysis can clearly identify the degree of influence of the occlusion on the steel stamp character, thereby helping to determine whether the recognition result is distorted due to the influence of the obstruction. The influence unit combines the overlap factor Ccyz obtained in the regional intersection analysis unit and the information of the rectangular frame establishment unit, and calculates the recognition influence coefficient Syxs after dimensionless processing. This coefficient can further comprehensively consider regional differences and depth information, and comprehensively evaluate the difficulty of steel stamp character recognition. The comparison unit compares the recognition influence coefficient Syxs with the preset evaluation threshold P to determine the current state of steel stamp character recognition. The specific processing is as follows: If the recognition influence coefficient Syxs ≥ evaluation threshold P, the system determines that there is an abnormality in the recognition, and emits three consecutive red identification lights to remind the operator that there may be a problem. If the recognition influence coefficient Syxs < evaluation threshold P, the system determines that the recognition is normal and emits a green identification light once, indicating that the drug status is normal. The evaluation unit takes corresponding sorting measures according to the color of the identification light: three consecutive red identification lights indicate that the drugs need to be detained and sorted to the quality inspection area to ensure quality control; a green identification light indicates that the drugs are normal and do not need to be detained, and the next inspection process will be continued. In summary, the system achieves efficient management of the steel stamp character recognition of the outer packaging of drugs through precise occlusion impact analysis, comprehensive impact coefficient calculation, and intelligent feedback and processing mechanism. This not only improves the accuracy of recognition, but also optimizes the processing process of abnormal drugs, ensuring the reliability of drug quality and detection efficiency.

[0115] Example 7

[0116] Please refer to Figure 1 and Figure 2 Specifically: A method for recognizing steel stamp characters comprises the following steps:

[0117] Step 1: Use visual monitoring equipment to capture images of the outer packaging of the drug to obtain an orthographic projection image, and use a 3D scanner to monitor the surface of the outer packaging of the drug to construct a relevant three-dimensional point cloud data set;

[0118] Step 2: preprocess the orthographic projection image to remove noise, smooth the image, perform geometric correction, and adjust the image quality based on image enhancement technology;

[0119] Step 3: Identify the difference areas based on the relevant 3D point cloud data set, set corresponding rectangular frames for several groups of difference areas, and identify the rectangular frames with the steel stamp characters based on the 3D point cloud data in each group of rectangular frames;

[0120] Step 4: Divide the rectangular frame where the stamped characters are identified into several groups of image areas, and monitor the reflection characteristics to obtain the reflectivity Fsl in each image area. According to the reflectivity Fsl in each image area, construct the reflection contrast Fbdd between different image areas. If the reflection contrast Fbdd exceeds the threshold, issue a cover recognition instruction.

[0121] Step 5: After receiving the cover recognition instruction, the overlap factor Ccyz between the area in the rectangular box where the steel stamp characters are identified in the recognition module and the cover area is analyzed, and the recognition influence coefficient Syxs is constructed in combination with the relevant three-dimensional point cloud data set;

[0122] Step 6: Pre-set the evaluation threshold P, and compare and analyze the evaluation threshold P with the recognition influence coefficient Syxs to estimate the difficulty of recognizing the steel stamp characters on the outer packaging of the current drug.

[0123] 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 steel stamp character recognition system, characterized in that: It includes an image capture module, an image preprocessing module, a recognition module, an occlusion analysis module, a comprehensive recognition analysis module and a feedback module; The image capture module is used to capture images of the outer packaging of the medicine using a visual monitoring instrument to obtain an orthographic projection image, and to monitor the surface of the outer packaging of the medicine in combination with a 3D scanner to construct a relevant three-dimensional point cloud data set; The image preprocessing module is used to preprocess the orthographic projection image, remove noise, smooth the image, perform geometric correction, and adjust the image quality based on image enhancement technology; The recognition module is used to identify the difference area based on the relevant three-dimensional point cloud data set, and set corresponding rectangular frames for several groups of difference areas, and identify the rectangular frame with the steel stamp characters according to the three-dimensional point cloud data in each group of rectangular frames; The recognition module includes a rectangular frame establishment unit and a steel stamp character area determination unit; The rectangular frame establishment unit establishes a plane coordinate system based on the plane where the drug outer packaging is located, and identifies the depth area based on the relevant three-dimensional point cloud data set obtained by the 3D scanner, wherein the relevant three-dimensional point cloud data set includes the position coordinates of each point on the surface of the drug outer packaging, the density of the points, the acquisition timestamp and the unique identifier of each point; at the same time, the distance from the 3D scanner to the plane coordinate system is set as a standard value. If the distance from the 3D scanner to the corresponding point on the surface of the drug outer packaging is greater than the standard value, it is preliminarily determined that the area corresponding to the corresponding point is an abnormal area, and corresponding rectangular frames are set for several groups of different areas to obtain the number of different areas Nys; The stamp character area determination unit is used to determine the number of points greater than the standard value in each group of rectangular frames according to the three-dimensional point cloud data in each group of rectangular frames, and take the rectangular frame with the largest number of points as the rectangular frame where the stamp character exists; The occlusion analysis module is used to divide the rectangular frame where the steel-printed characters are identified into several groups of image areas, and monitor the reflection characteristics to obtain the reflectivity Fsl in each image area, and construct the reflection contrast Fbdd between different image areas based on the reflectivity Fsl in each image area. If the reflection contrast Fbdd exceeds the threshold, a cover recognition instruction is issued externally; The comprehensive recognition and analysis module is used to analyze the overlap factor Ccyz between the area in the rectangular frame where the steel stamp characters exist and the masked area after receiving the mask recognition instruction, and to construct the recognition influence coefficient Syxs in combination with the relevant three-dimensional point cloud data set; The comprehensive identification and analysis module includes a regional cross-analysis unit and an impact unit; The area intersection analysis unit is used to analyze the overlap factor Ccyz between the area in the rectangular frame where the steel stamp characters exist and the cover area after receiving the cover recognition instruction, which is specifically obtained according to the following formula: In the formula, A represents the area of ​​the blocked area, B represents the area of ​​the stamped character area, ∩ represents a logical "and" operation, and ∪ represents a logical "or" operation; The influence unit is used to establish a unit based on a rectangular frame, and then combined with the overlap degree factor Ccyz obtained in the regional cross analysis unit, and after dimensionless processing, the identification influence coefficient Syxs is obtained, which is specifically obtained according to the following formula: Where Nys represents the number of difference regions, Scc represents the depth difference, α and β are weight values, and C is the correction constant; The feedback module is used to pre-set an evaluation threshold value P, and compare and analyze the evaluation threshold value P with the recognition influence coefficient Syxs to estimate the difficulty of recognizing the steel-printed characters in the current drug outer packaging.

2. A steel stamp character recognition system according to claim 1, characterized in that: The image capture module includes a capture unit and a three-dimensional state acquisition unit; The capture unit is used to perform orthographic projection photography of the outer packaging of the drug by using a visual monitoring instrument, wherein the visual monitoring instrument includes a GigE industrial camera and a network camera; The three-dimensional state acquisition unit is used to scan the degree of depression on the surface of the outer packaging of the medicine using 3D scanning technology, obtain three-dimensional point cloud data of the surface of the outer packaging of the medicine by emitting a laser from the 3D scanner and receiving the reflected signal, and use point cloud registration technology to splice and integrate the three-dimensional point cloud data of multiple scans to construct a related three-dimensional point cloud data set.

3. A steel stamp character recognition system according to claim 1, characterized in that: The image preprocessing module includes a processing unit and an enhancement unit; The processing unit is used to use wavelet denoising technology to remove noise components in the wavelet domain, and use histogram equalization to adjust the brightness, and compress the image data in PNG format; The enhancement unit is used to remap the gray value of each pixel in the image by using the image enhancement technology and CDF.

4. A steel stamp character recognition system according to claim 1, characterized in that: The occlusion analysis module includes a data testing unit, a reflection analysis unit, an early warning unit and a differentiation unit; The data testing unit is used to divide the rectangular frame where the steel-printed characters are identified into a plurality of groups of image areas, and monitor the reflection characteristics of the plurality of groups of image areas using a spectroscopic reflectometer to obtain the reflectivity Fsl of all pixels in each image area; The reflection analysis unit includes a local analysis subunit and an in-frame overall analysis subunit; The local analysis subunit is used to obtain the mean reflectivity in each image area by combining a statistical averaging algorithm. The mean reflectivity in each image area Obtained by the following formula: Where i = 1, 2, 3, ..., n, n represents the total number of pixels in the image area, Fsl i Expressed as the reflectance of each pixel in the image area.

5. A steel stamp character recognition system according to claim 4, characterized in that: The frame overall analysis subunit is used to analyze the average reflectivity of each image area. Calculate the reflection contrast Fbdd between different areas, which can be obtained as follows: The significance of this formula is to reflect the area where the reflection characteristics of different regions vary, where: and are the mean reflectance values ​​of two different image regions; The warning unit is used to pre-set a threshold value, and compares and analyzes the reflection contrast Fbdd between different areas with the threshold value to determine whether there is a reflection difference caused by obstacles in the rectangular frame where the steel-printed characters are stored. The specific comparison content is as follows: If the reflection contrast Fbdd exceeds the threshold, it is determined that there is a reflection difference caused by obstacles in the rectangular frame containing the stamped characters, and a cover recognition instruction will be issued. If the reflection contrast Fbdd does not exceed the threshold, it is determined that there is no reflection difference caused by obstacles in the rectangular frame where the stamped characters are stored; The differentiation unit is used to automatically determine the adaptation threshold using the Otsu algorithm, perform threshold processing on several groups of image areas, divide the image area into two parts, and take the area where the reflectivity Fsl of each pixel is greater than the threshold as the foreground, and take the area where the reflectivity Fsl of each pixel is less than the threshold as the background. Combined with the characteristics of the foreground, the area where the foreground is located is regarded as the obstacle occlusion area.

6. A steel stamp character recognition system according to claim 1, characterized in that: The feedback module includes a comparison unit and an estimation unit; The comparison unit is used to compare and analyze the recognition influence coefficient Syxs with the evaluation threshold P to estimate the difficulty of recognizing the steel stamp characters in the current drug outer packaging. The specific comparison content is as follows: If the recognition influence coefficient Syxs ≥ the evaluation threshold P, it means that the recognition of the steel stamp characters in the current drug outer packaging is at an abnormal level, and the red identification light will be emitted three times in succession; If the recognition influence coefficient Syxs is less than the evaluation threshold value P, it means that the recognition of the steel stamp characters in the current drug outer packaging is at a normal level, and a green identification light will be emitted once; The evaluation unit is used to receive identification lights and take corresponding sorting operations according to identification lights of different colors. When a red identification light is received three times in a row, the corresponding drugs will be detained and classified into the area for quality inspection; when a green identification light is received once, the corresponding drugs will not be detained and the next inspection process will be continued.

7. A method for recognizing steel stamp characters, applied to a steel stamp character recognition system according to any one of claims 1 to 6, characterized in that: The following steps are included: Step 1: Use visual monitoring equipment to capture images of the outer packaging of the drug to obtain an orthographic projection image, and use a 3D scanner to monitor the surface of the outer packaging of the drug to construct a relevant three-dimensional point cloud data set; Step 2: preprocess the orthographic projection image to remove noise, smooth the image, perform geometric correction, and adjust the image quality based on image enhancement technology; Step 3: Identify the difference areas based on the relevant 3D point cloud data set, set corresponding rectangular frames for several groups of difference areas, and identify the rectangular frames with the steel stamp characters based on the 3D point cloud data in each group of rectangular frames; Step 4: Divide the rectangular frame where the stamped characters are identified into several groups of image areas, and monitor the reflection characteristics to obtain the reflectivity Fsl in each image area. According to the reflectivity Fsl in each image area, construct the reflection contrast Fbdd between different image areas. If the reflection contrast Fbdd exceeds the threshold, issue a cover recognition instruction. Step 5: After receiving the cover recognition instruction, the overlap factor Ccyz between the area in the rectangular box where the steel stamp characters are identified in the recognition module and the cover area is analyzed, and the recognition influence coefficient Syxs is constructed in combination with the relevant three-dimensional point cloud data set; Step 6: Pre-set the evaluation threshold P, and compare and analyze the evaluation threshold P with the recognition influence coefficient Syxs to estimate the difficulty of recognizing the steel stamp characters on the outer packaging of the current drug.