A method and system for identifying surgical sutures based on an industrial camera system

Abstract

This invention discloses a method and system for identifying surgical sutures based on an industrial camera system, belonging to the field of surgical suture quality inspection technology. It includes a transparent conveyor belt, an upper magnifying device and a lower magnifying device located above and below the transparent conveyor belt, an upper industrial camera located above the transparent conveyor belt and the upper magnifying device, a lower industrial camera located below the transparent conveyor belt and the lower magnifying device, and a central processing system connected to the upper and lower industrial cameras respectively. The central processing system is connected to an image processing unit, a 3D image synthesis unit, and a key judgment point selection unit. This invention can obtain a complete panoramic view of the suture shape and fibrous distribution, truly reflecting the overall fibrous level and the overall surface quality. Anomaly detection does not require the sample to remain under the lens for an extended period, does not affect the conveyor belt speed, significantly improves detection efficiency, and is suitable for continuous batch production lines.

Description

Technical Field

[0001] This invention relates to the field of surgical suture quality inspection technology, specifically to a method and system for identifying surgical sutures based on an industrial camera system. Background Technology

[0002] Surgical sutures are typically made from multiple strands of fine monofilaments through melting, spinning, and weaving. The materials are either absorbable or non-absorbable. In actual production, due to machine errors and static electricity between the strands, the filaments can break and detach, forming fuzz on the suture surface. Fuzz significantly impacts downstream weaving and dyeing processes, directly determining the quality of the final product. Therefore, fuzz detection is one of the crucial quality indicators for surgical sutures.

[0003] For example, Chinese patent applications CN201611036029.0 and CN201610495406.0 disclose a method for detecting the quality of silk threads using image processing technology. A CCD camera is used to capture images of the sample, and then an image processing system is used to ultimately filter out areas that differ from normal silk threads. While this method can achieve the screening of silk threads at a low cost, it has the following drawbacks:

[0004] 1. The fine lines on the back of the camera are not easily identified, making it impossible to obtain a complete panoramic photo of the surgical sutures; it also fails to demonstrate the overall level of fine lines, which has limited use for manufacturers in subsequent process improvements.

[0005] 2. The sample needs to remain under the lens for a long time to ensure that a very suitable test image is obtained, resulting in low testing efficiency.

[0006] Based on this, the present invention designs a method and system for identifying surgical sutures based on an industrial camera system to solve the above problems. Summary of the Invention

[0007] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a method and system for identifying surgical sutures based on an industrial camera system.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A system for recognizing surgical sutures based on an industrial camera system includes a transparent conveyor belt, an upper magnifying device and a lower magnifying device located on the upper and lower sides of the transparent conveyor belt respectively, an upper industrial camera located above the transparent conveyor belt and the upper magnifying device, a lower industrial camera located below the transparent conveyor belt and the lower magnifying device, and a central processing system connected to the upper industrial camera and the lower industrial camera respectively.

[0010] The central processing system is connected to the barcode scanning module. Each surgical suture to be tested has a unique sample number. During testing, the outer packaging of the suture is scanned by the barcode scanning module. The suture is laid flat on a transparent conveyor belt. The upper and lower magnifying devices magnify the upper and lower surfaces of the suture on the transparent conveyor belt, respectively. The upper and lower industrial cameras take pictures of the magnified upper and lower surfaces of the suture, respectively. All the pictures are transmitted to the central processing system.

[0011] The central processing system is connected to an image processing unit, which is used to perform noise reduction processing, binarization segmentation processing, and anomaly recognition algorithm on the image.

[0012] The binarization segmentation method is as follows:

[0013] (1) Given that the image pixel size is m×n, calculate the initial threshold T0:

[0014]

[0015] Where f(x,y) is the gray value of pixel (x,y), and Q is the total number of pixels = m×n;

[0016] (2) Divide the image into background C0 and foreground C1 according to the initial threshold T0, and calculate the pixel probability of the two types of regions:

[0017]

[0018] Where, p i =q i / Q, q i Let be the total number of pixels with grayscale value i, and P0 and P1 be the probabilities of C0 and C1, respectively.

[0019] (3) Calculate the expected gray values ​​of the two types of regions and the overall expected gray value:

[0020]

[0021] Where μ0 is the expected gray value of C0, μ1 is the expected gray value of C1, and μ is the overall expectation;

[0022] (4) Maximize the inter-class variance σ B ², Determine the optimal threshold T:

[0023] ;

[0024] The anomaly detection method is as follows:

[0025] (1) Extract the region of interest (ROI) containing only the suture target;

[0026] (2) Scan the ROI according to the preset parameters and mark the areas with abnormal fuzz, abnormal thread ends, and abnormal thread diameter deviation;

[0027] The central processing system is connected to a 3D image synthesis unit. The 3D image synthesis unit uses the upper surface image A and the lower surface image B of the same suture, which have been processed by the image processing unit, as surface texture maps and attaches them to the surface of a virtual cylinder model with the same diameter as the suture being measured.

[0028] The central processing system is connected to a key decision point selection unit, which is used to automatically select key decision points along the suture extension direction on the reconstructed 3D suture model; at the decision point location, based on the abnormal results identified and marked by the image processing unit, it determines whether there are abnormalities such as fuzz, thread ends, or diameter deviations at each decision point, and determines whether the sample is qualified.

[0029] Furthermore, the upper and lower industrial cameras are arranged in a staggered manner.

[0030] Furthermore, the noise reduction process uses a Gaussian function to reduce lighting interference, as shown in the formula:

[0031]

[0032] Where G(x, y) is the Gaussian kernel function value, the Gaussian kernel is 3×3 or 5×5, x, y are pixel coordinates, σ is the standard deviation, and σ takes a value of 0.5-1.0.

[0033] Furthermore, the preset parameters are set as follows:

[0034] If the length of the fuzz is ≥0.5mm, it is judged as a fuzz defect;

[0035] If the diameter of the thread end is ≥0.1mm, it is considered a thread end defect;

[0036] If the diameter deviation is ≥0.05mm, it is judged as a diameter unevenness defect.

[0037] Furthermore, based on the OTSU binarization segmentation results, the binary image is first processed for edge optimization, and then the image is divided into foreground and background. The area with a gray value greater than or equal to the optimal threshold is the suture foreground, and the rest is the conveyor belt background. Then, the minimum bounding rectangle of the foreground target is obtained through contour detection. The original image is cropped with this rectangle as the boundary to remove irrelevant background areas, resulting in a foreground region of interest (ROI) containing only the suture target.

[0038] Furthermore, the judgment rule is as follows: if there are no abnormalities at any judgment point, the system determines that the suture sample is qualified; if there is an abnormality at any judgment point, the system determines that the suture sample is unqualified.

[0039] Furthermore, the central processing system is also connected to a data retention unit and a cloud synchronization unit; the data retention unit stores the "sample number, A-figure, B-figure, 3D figure and sample judgment result" data to the local hard disk; the cloud synchronization unit synchronizes the above data to the cloud server.

[0040] Furthermore, the central processing system automatically calculates the pass rate for a single batch, using the following formula: Pass rate = Number of qualified samples / Total number of samples × 100%.

[0041] To better achieve the objectives of this invention, the present invention also provides a method for identifying surgical sutures based on an industrial camera system. The method for identifying surgical sutures based on an industrial camera system includes the following steps:

[0042] 1. Sample number entry: Scan the barcode or QR code on the outer packaging of the surgical suture to be tested using the barcode recognition module to obtain the unique sample number corresponding to the suture, and transmit the sample number to the central processing system.

[0043] II. Thread Laying and Image Acquisition: Remove the outer packaging of the thread and lay it flat on the transparent conveyor belt; start the transparent conveyor belt to move the thread, and use the upper and lower magnification devices located on the upper and lower sides of the transparent conveyor belt to magnify the upper and lower surfaces of the thread respectively. The upper industrial camera above acquires image A of the upper surface of the thread, and the lower industrial camera below acquires image B of the lower surface of the thread; number images A and B according to the sample number and transmit them synchronously to the central processing system;

[0044] Third, perform noise reduction, binarization segmentation, and anomaly detection algorithms on the images;

[0045] IV. 3D suture model synthesis: The central processing system uses the 3D image synthesis unit to use the processed images A and B as texture maps and attach them to the surface of a virtual cylindrical model with the same diameter as the suture being measured.

[0046] V. Selection of Key Judgment Points and Acceptance Judgment: On the reconstructed 3D suture model, key judgment points are automatically selected along the suture length direction; based on the abnormal results identified and marked in step three, it is determined whether there are abnormalities such as fuzz, thread ends, or diameter deviations at each judgment point;

[0047] If all the judgment points are normal, the suture sample is deemed qualified; if any judgment point is abnormal, it is deemed unqualified.

[0048] VI. Data storage and synchronization.

[0049] Furthermore, step three specifically involves:

[0050] The central processing system performs noise reduction processing on the acquired images A and B through the image processing unit. The noise reduction process reduces lighting interference through a Gaussian function.

[0051] The image is then subjected to binarization segmentation. The binarization segmentation method is as follows:

[0052] (1) Assume the image pixel size is m×n and the gray level L=256. First, calculate the initial threshold T0:

[0053]

[0054] Where f(x,y) is the gray value of pixel (x,y), and Q is the total number of pixels = m×n;

[0055] (2) Divide the image into background C0 and foreground C1 according to the initial threshold T0, and calculate the pixel probability of the two types of regions:

[0056]

[0057] Where, p i =q i / Q, q i Let be the total number of pixels with grayscale value i, and P0 and P1 be the probabilities of C0 and C1, respectively.

[0058] (3) Calculate the expected gray values ​​of the two types of regions and the overall expected gray value:

[0059]

[0060] Where μ0 is the expected gray value of C0, μ1 is the expected gray value of C1, and μ is the overall expectation;

[0061] (4) Maximize the inter-class variance σ B ², Determine the optimal threshold T:

[0062]

[0063] Anomaly identification was then performed using the following method: Based on the OTSU binarization segmentation results, edge optimization was first performed on the binary image. Then, the image was divided into foreground and background, with the area where the gray value was greater than or equal to the optimal threshold being the foreground of the seam, and the rest being the background of the conveyor belt. Next, the minimum bounding rectangle of the foreground target was obtained through contour detection. The original image was cropped using this rectangle as the boundary to remove irrelevant background areas, resulting in a region of interest (ROI) containing only the seam target. Within the extracted foreground ROI, three types of anomalies were determined: fuzz anomaly determination: if the fuzz length is ≥0.5mm, it is determined to be a fuzz defect, and the corresponding area is marked; thread end anomaly determination: if the thread end diameter is ≥0.1mm, it is determined to be a thread end defect, and the corresponding area is marked; seam diameter deviation anomaly determination: if the diameter deviation is ≥0.05mm, it is determined to be a diameter unevenness defect, and the area with deviation exceeding the limit is marked.

[0064] Compared with the prior art, the beneficial effects of this invention are as follows: This invention uses a transparent conveyor belt in conjunction with industrial cameras arranged in a staggered manner to simultaneously acquire images A and B of the upper and lower surfaces of the seam; and through 3D cylindrical model reconstruction, the upper and lower surface images are stitched together to form a complete 360° three-dimensional model of the seam, which can identify defects on the upper and lower surfaces, and can obtain a complete panoramic view of the seam shape and fuzz distribution, truly reflecting the overall fuzz level and the overall surface quality.

[0065] This invention employs a transparent conveyor belt, upper and lower magnification devices, and an industrial camera, along with a fast algorithm for ROI extraction and global anomaly recognition, to achieve anomaly detection of seams. It eliminates the need for samples to remain under the lens for extended periods, does not affect the conveyor belt's operating speed, significantly improves detection efficiency, and is suitable for continuous batch production lines. Attached Figure Description

[0066] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0067] Figure 1 This is a flowchart of a method for identifying surgical sutures based on an industrial camera system according to the present invention;

[0068] Figure 2 This is a system structure diagram of a surgical suture recognition system based on an industrial camera system according to the present invention.

[0069] The labels in the diagram represent: 1. Upper industrial camera; 2. Lower industrial camera; 3. Transparent conveyor belt; 4. Upper magnification device; 5. Lower magnification device; 6. Central processing system. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0071] The terms "left," "right," "front," "back," "up," and "down" used in the following description refer to the orientation from the perspective of the front view.

[0072] Example 1: Please refer to the accompanying drawings in the instruction manual. Figure 2 A system for identifying surgical sutures based on an industrial camera system includes a transparent conveyor belt 3, an upper magnifying device 4 and a lower magnifying device 5 located above and below the transparent conveyor belt 3 respectively, an upper industrial camera 1 located above the transparent conveyor belt 3 and the upper magnifying device 4, a lower industrial camera 2 located below the transparent conveyor belt 3 and the lower magnifying device 5, and a central processing system 6 connected to the upper industrial camera 1 and the lower industrial camera 2 respectively. The central processing system 6 includes a central processing unit (CPU).

[0073] Each surgical suture to be tested (hereinafter referred to as suture) has a unique sample number, such as "suture-20250101-001"; correspondingly, the outer packaging of the suture is equipped with a barcode or QR code that corresponds one-to-one with the unique sample number.

[0074] During testing, the outer packaging of the sewing thread is first scanned using the set barcode scanning module to quickly and accurately read the sample number. The barcode scanning module is connected to the central processing system 6.

[0075] Afterwards, the outer packaging of the sewing thread is removed, and the sewing thread is laid flat on the transparent conveyor belt 3. The upper magnifying device 4 and the lower magnifying device 5 are used to magnify the upper and lower surfaces of the sewing thread on the transparent conveyor belt 3 respectively. The upper industrial camera 1 and the lower industrial camera 2 are used to take pictures of the magnified upper and lower surfaces of the sewing thread respectively. All the pictures are transmitted to the central processing system 6.

[0076] The transparent conveyor belt 3 is started, and it moves the sewing thread to the right. First, the upper surface image A of the sewing thread is captured by the upper industrial camera 1, and the image is numbered according to the sample number, for example, 001-A-001. Then, the lower surface image B of the sewing thread is captured by the lower industrial camera 2, and the image is numbered according to the sample number, for example, 001-B-001. The upper surface image A, the lower surface image B and their numbers are all transmitted to the central processing system 6.

[0077] Preferably, the upper industrial camera 1 and the lower industrial camera 2 can be arranged in a staggered manner;

[0078] The central processing system 6 is connected to an image processing unit, which is used to perform noise reduction processing, binarization segmentation processing, and anomaly recognition algorithm on the image; the image processing unit adopts an image processor.

[0079] Specifically, noise reduction uses a Gaussian function to reduce lighting interference, as shown in the formula:

[0080]

[0081] Where G(x, y) is the Gaussian kernel function value, the Gaussian kernel is 3×3 or 5×5, x, y are pixel coordinates, σ is the standard deviation, and σ takes a value of 0.5-1.0;

[0082] The binarization segmentation (OTSU method) method is as follows:

[0083] (1) Assume the image pixel size is m×n (e.g., 1920×1080), and the gray level L=256 (0-255). First, calculate the initial threshold T0:

[0084]

[0085] Where f(x,y) is the gray value of pixel (x,y), and Q is the total number of pixels = m×n;

[0086] (2) Divide the image into background C0 (grayscale value < T0, corresponding to transparent conveyor belt 3) and foreground C1 (grayscale value ≥ T0, corresponding to seam) according to the initial threshold T0, and calculate the pixel probability of the two types of regions:

[0087]

[0088] Where, p i =q i / Q, q i Let be the total number of pixels with grayscale value i, and P0 and P1 be the probabilities of C0 and C1, respectively.

[0089] (3) Calculate the expected gray values ​​of the two types of regions and the overall expected gray value:

[0090]

[0091] Where μ0 is the expected gray value of C0, μ1 is the expected gray value of C1, and μ is the overall expectation;

[0092] (4) Maximize the inter-class variance σ B ², Determine the optimal threshold T (at which the segmentation effect is optimal):

[0093]

[0094] The anomaly detection method is as follows:

[0095] (1) Extract the foreground ROI;

[0096] Based on the OTSU binarization segmentation results, edge optimization processing is first performed on the binary image: erosion and dilation operations are performed sequentially using 3×3 rectangular structuring elements to remove isolated noise points and edge burrs, and to complete the broken seam contours, making the edges continuous and smooth, thus avoiding interference with subsequent contour detection.

[0097] The image is then divided into foreground and background. The area with a gray value greater than or equal to the optimal threshold is the suture foreground, and the rest is the conveyor belt background.

[0098] Then, the minimum bounding rectangle of the foreground target is obtained through contour detection. The original image is cropped using this rectangle as the boundary to remove irrelevant background areas, resulting in a region of interest (ROI) that contains only the seam target. This achieves focused detection range and improves the accuracy and efficiency of subsequent anomaly recognition.

[0099] (2) Scan the ROI according to the preset parameters (fiber length ≥ 0.5 mm, thread diameter ≥ 0.1 mm, suture diameter deviation ≥ 0.05 mm) and mark the abnormal area.

[0100] Within the extracted foreground ROI, three types of anomalies are identified: Frayed thread anomaly: If the fringe length is ≥0.5mm, it is identified as a fringe defect, and the corresponding area is marked. Thread end anomaly: If the thread end diameter is ≥0.1mm, it is identified as a thread end defect, and the corresponding area is marked. Thread diameter deviation anomaly: If the diameter deviation is ≥0.05mm, it is identified as a diameter unevenness defect, and the area exceeding the deviation limit is marked. The anomaly areas are then visually marked.

[0101] Furthermore, camera calibration is performed in advance using standard gauge blocks to establish a mapping relationship between pixels and actual physical dimensions (mm).

[0102] The central processing system 6 is connected to a 3D image synthesis unit. The 3D image synthesis unit uses the upper surface image A and lower surface image B of the same suture, processed by the image processing unit, as surface texture maps and applies them to the surface of a virtual cylindrical model with the same diameter as the suture being measured. The model can be rotated 360° by dragging and sliding with the mouse to visually observe the details such as the texture, fuzz, thread ends, and diameter fluctuations of the upper, lower, and side surfaces of the suture.

[0103] For example, if the nominal diameter of the suture being tested is 0.3mm, the system will automatically match and load a standard virtual cylinder model with a diameter of 0.3mm, and map the two two-dimensional suture images, upper surface image A and lower surface image B, to the corresponding curved surface positions of the cylinder model to reconstruct a three-dimensional visualization model of the suture.

[0104] Furthermore, the upper surface image A is mapped onto the upper half of the virtual cylinder, and the lower surface image B is mapped onto the lower half of the virtual cylinder, and the two images are then stitched together to form a complete 3D model of the suture.

[0105] The central processing system 6 is connected to a key decision point selection unit. The key decision point selection unit is used to automatically select 5 key decision points (small rectangular areas of 1cm×1cm centered on the decision points) along the extension direction of the suture on the reconstructed 3D suture model. The positions are as follows: 1cm from the beginning of the suture, 12.5cm from the beginning of the suture, 25cm from the middle of the suture, 12.5cm from the end of the suture, and 1cm from the end of the suture. At the above 5 decision point positions, based on the abnormal results identified and marked by the image processing unit, it is determined whether there are abnormalities such as fuzz, thread ends, or diameter deviations at each decision point, and the sample is judged as qualified.

[0106] The specific judgment rule is as follows: if there are no abnormalities in any of the 5 judgment points, the system determines that the suture sample is qualified; if there is an abnormality in any one of the judgment points, the system determines that the suture sample is unqualified.

[0107] The central processing system 6 is also connected to a data retention unit and a cloud synchronization unit; the data retention unit stores the "sample number, A-figure, B-figure, 3D figure and sample judgment result" data to the local hard disk; the cloud synchronization unit synchronizes the above data to the cloud server, which can realize real-time viewing by the monitoring terminal outside the workshop;

[0108] The central processing system 6 can also automatically calculate the pass rate of a single batch. The calculation method is: pass rate = number of qualified samples / total number of samples × 100%.

[0109] Example 2: Figure 1 As shown, a method for identifying surgical sutures based on an industrial camera system includes the following steps:

[0110] 1. Sample Number Entry: Scan the barcode or QR code on the outer packaging of the surgical suture to be tested using the barcode scanning module to obtain the unique sample number corresponding to the suture, and then transmit the sample number to the central processing system.

[0111] II. Thread Laying and Image Acquisition: Remove the outer packaging of the thread and lay it flat on the transparent conveyor belt; start the transparent conveyor belt to move the thread, and use the upper and lower magnification devices located on the upper and lower sides of the transparent conveyor belt to magnify the upper and lower surfaces of the thread respectively. The upper industrial camera above acquires image A of the upper surface of the thread, and the lower industrial camera below acquires image B of the lower surface of the thread; number images A and B according to the sample number and transmit them synchronously to the central processing system.

[0112] Third, the image undergoes noise reduction, binarization segmentation, and anomaly detection algorithm.

[0113] The central processing system performs noise reduction processing on the acquired images A and B through the image processing unit. The noise reduction process reduces lighting interference through a Gaussian function.

[0114] The image is then subjected to binarization segmentation. The binarization segmentation (OTSU method) method is as follows:

[0115] (1) Assume the image pixel size is m×n (e.g., 1920×1080), and the gray level L=256 (0-255). First, calculate the initial threshold T0:

[0116]

[0117] Where f(x,y) is the gray value of pixel (x,y), and Q is the total number of pixels = m×n;

[0118] (2) Divide the image into background C0 (grayscale value < T0, corresponding to transparent conveyor belt 3) and foreground C1 (grayscale value ≥ T0, corresponding to seam) according to the initial threshold T0, and calculate the pixel probability of the two types of regions:

[0119]

[0120] Where, p i =q i / Q, q i Let be the total number of pixels with grayscale value i, and P0 and P1 be the probabilities of C0 and C1, respectively.

[0121] (3) Calculate the expected gray values ​​of the two types of regions and the overall expected gray value:

[0122]

[0123] Where μ0 is the expected gray value of C0, μ1 is the expected gray value of C1, and μ is the overall expectation;

[0124] (4) Maximize the inter-class variance σ B ², Determine the optimal threshold T:

[0125]

[0126] Then, anomaly detection is performed, using the following method:

[0127] (1) Extract the foreground ROI;

[0128] Based on the OTSU binarization segmentation results, edge optimization processing is first performed on the binary image: erosion and dilation operations are performed sequentially using 3×3 rectangular structuring elements to remove isolated noise points and edge burrs, and to complete the broken seam contours, making the edges continuous and smooth, thus avoiding interference with subsequent contour detection.

[0129] The image is then divided into foreground and background. The area with a gray value greater than or equal to the optimal threshold is the suture foreground, and the rest is the conveyor belt background.

[0130] Then, the minimum bounding rectangle of the foreground target is obtained through contour detection. The original image is cropped using this rectangle as the boundary to remove irrelevant background areas, resulting in a region of interest (ROI) that contains only the seam target. This achieves focused detection range and improves the accuracy and efficiency of subsequent anomaly recognition.

[0131] (2) Scan the ROI according to the preset parameters (fiber length ≥ 0.5 mm, thread diameter ≥ 0.1 mm, suture diameter deviation ≥ 0.05 mm) and mark the abnormal area.

[0132] Within the extracted foreground ROI, three types of anomalies are identified: Frayed thread anomaly: If the fringe length is ≥0.5mm, it is identified as a fringe defect, and the corresponding area is marked. Thread end anomaly: If the thread end diameter is ≥0.1mm, it is identified as a thread end defect, and the corresponding area is marked. Thread diameter deviation anomaly: If the diameter deviation is ≥0.05mm, it is identified as a diameter unevenness defect, and the area exceeding the deviation limit is marked. The anomaly areas are then visually marked.

[0133] IV. 3D suture model synthesis: The central processing system uses the 3D image synthesis unit to process images A and B as texture maps and attach them to the surface of a virtual cylinder model with the same diameter as the suture being measured. The upper surface image A is mapped to the upper half of the cylinder, and the lower surface image B is mapped to the lower half of the cylinder, generating a 3D visualization model of the suture that can be rotated 360°.

[0134] V. Selection of Key Judgment Points and Acceptance Judgment: On the reconstructed 3D suture model, five key judgment points are automatically selected along the suture length direction: 1cm from the beginning, 12.5cm from the beginning, 25cm from the middle, 12.5cm from the end, and 1cm from the end; Based on the abnormal results identified and marked in step three, it is determined whether there are any abnormalities such as fuzz, thread ends, or diameter deviations at each judgment point.

[0135] If all five judgment points are normal, the suture sample is deemed acceptable; if any one judgment point is abnormal, it is deemed unacceptable.

[0136] VI. Data Storage and Synchronization: The central processing system stores the sample number, image A, image B, 3D model, and pass / fail judgment results in the local data retention unit and uploads them to the cloud server through the cloud synchronization unit for remote viewing; at the same time, it automatically counts the total number of tests and the number of qualified tests for a single batch, calculates and outputs the batch pass rate, and displays it on an external monitor.

[0137] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A system for recognizing surgical sutures based on an industrial camera system, characterized in that, It includes a transparent conveyor belt (3), an upper magnifying device (4) and a lower magnifying device (5) located on the upper and lower sides of the transparent conveyor belt (3) respectively, an upper industrial camera (1) located above the transparent conveyor belt (3) and the upper magnifying device (4), a lower industrial camera (2) located below the transparent conveyor belt (3) and the lower magnifying device (5), and a central processing system (6) connected to the upper industrial camera (1) and the lower industrial camera (2) respectively. The central processing system (6) is connected to the barcode scanning module. Each surgical suture to be tested has a unique sample number. During testing, the outer packaging of the suture is scanned by the barcode scanning module. The suture is laid flat on the transparent conveyor belt (3). The upper magnification device (4) and the lower magnification device (5) magnify the upper and lower surfaces of the suture on the transparent conveyor belt (3) respectively. The upper industrial camera (1) and the lower industrial camera (2) take pictures of the magnified upper and lower surfaces of the suture respectively. The pictures are transmitted to the central processing system (6). The central processing system (6) is connected to an image processing unit, which is used to perform noise reduction processing, binarization segmentation processing, and anomaly recognition algorithm on the image. The binarization segmentation method is as follows: (1) Given that the image pixel size is m×n, calculate the initial threshold T0: ； Where f(x,y) is the gray value of pixel (x,y), and Q is the total number of pixels = m×n; (2) Divide the image into background C0 and foreground C1 according to the initial threshold T0, and calculate the pixel probability of the two types of regions: ； Where, p i =q i / Q, q i Let be the total number of pixels with grayscale value i, and P0 and P1 be the probabilities of C0 and C1, respectively. (3) Calculate the expected gray values ​​of the two types of regions and the overall expected gray value: ； Where μ0 is the expected gray value of C0, μ1 is the expected gray value of C1, and μ is the overall expectation; (4) Maximize the inter-class variance σ B ², Determine the optimal threshold T: ； The anomaly detection method is as follows: (1) Extract the region of interest (ROI) containing only the suture target; (2) Scan the ROI according to the preset parameters and mark the areas with abnormal fuzz, abnormal thread ends, and abnormal thread diameter deviation; The central processing system (6) is connected to a 3D image synthesis unit. The 3D image synthesis unit uses the upper surface image A and the lower surface image B of the same suture after processing by the image processing unit as surface texture maps and attaches them to the surface of a virtual cylinder model with the same diameter as the suture being measured. The central processing system (6) is connected to a key decision point selection unit. The key decision point selection unit is used to automatically select key decision points along the suture extension direction on the reconstructed 3D suture model. At the decision point location, based on the abnormal results identified and marked by the image processing unit, it is determined whether there are abnormalities such as fuzz, thread ends, or diameter deviations at each decision point, and whether the sample is qualified is determined.

2. The method and system for identifying surgical sutures based on an industrial camera system according to claim 1, characterized in that, The upper industrial camera (1) and the lower industrial camera (2) are arranged in a staggered manner.

3. The method and system for identifying surgical sutures based on an industrial camera system according to claim 2, characterized in that, Noise reduction uses a Gaussian function to reduce lighting interference, the formula is: ； Where G(x, y) is the Gaussian kernel function value, the Gaussian kernel is 3×3 or 5×5, x, y are pixel coordinates, σ is the standard deviation, and σ takes a value of 0.5-1.

0.

4. The method and system for identifying surgical sutures based on an industrial camera system according to claim 3, characterized in that, The default parameters are set as follows: If the length of the fuzz is ≥0.5mm, it is judged as a fuzz defect; If the diameter of the thread end is ≥0.1mm, it is considered a thread end defect; If the diameter deviation is ≥0.05mm, it is judged as a diameter unevenness defect.

5. The method and system for identifying surgical sutures based on an industrial camera system according to claim 4, characterized in that, Based on the OTSU binarization segmentation results, the binary image is first processed for edge optimization, and then the image is divided into foreground and background. The area with gray value greater than or equal to the optimal threshold is the suture foreground, and the rest is the conveyor belt background. Then, the minimum bounding rectangle of the foreground target is obtained by contour detection. The original image is cropped with this rectangle as the boundary to remove irrelevant background areas and obtain the foreground region of interest (ROI) containing only the suture target.

6. The method and system for identifying surgical sutures based on an industrial camera system according to claim 5, characterized in that, The judgment rule is as follows: if there are no abnormalities at any judgment point, the system determines that the suture sample is qualified; if there is an abnormality at any judgment point, the system determines that the suture sample is unqualified.

7. The method and system for identifying surgical sutures based on an industrial camera system according to claim 6, characterized in that, The central processing system (6) is also connected to a data retention unit and a cloud synchronization unit; the data retention unit stores the "sample number, A figure, B figure, 3D figure and sample judgment result" data to the local hard disk; the cloud synchronization unit synchronizes the above data to the cloud server.

8. The method and system for identifying surgical sutures based on an industrial camera system according to claim 7, characterized in that, The central processing system (6) automatically calculates the pass rate of a single batch. The calculation method is: pass rate = number of qualified samples / total number of samples × 100%.

9. A method for identifying surgical sutures based on an industrial camera system, used in the method for identifying surgical sutures based on an industrial camera system as described in claim 8, characterized in that, Includes the following steps:

1. Sample number entry: Scan the barcode or QR code on the outer packaging of the surgical suture to be tested using the barcode recognition module to obtain the unique sample number corresponding to the suture, and transmit the sample number to the central processing system. II. Thread Laying and Image Acquisition: Remove the outer packaging of the thread and lay it flat on the transparent conveyor belt; start the transparent conveyor belt to move the thread, and use the upper and lower magnification devices located on the upper and lower sides of the transparent conveyor belt to magnify the upper and lower surfaces of the thread respectively. The upper industrial camera above acquires image A of the upper surface of the thread, and the lower industrial camera below acquires image B of the lower surface of the thread; number images A and B according to the sample number and transmit them synchronously to the central processing system; Third, perform noise reduction, binarization segmentation, and anomaly detection algorithms on the images; IV. 3D suture model synthesis: The central processing system uses the 3D image synthesis unit to use the processed images A and B as texture maps and attach them to the surface of a virtual cylindrical model with the same diameter as the suture being measured. V. Selection of key judgment points and qualification judgment: Key judgment points are automatically selected along the suture length on the reconstructed 3D suture model. Based on the abnormal results identified and marked in step three, determine whether there are any abnormalities such as lint, thread ends, or diameter deviations at each judgment point; If all the judgment points are normal, the suture sample is deemed qualified; if any judgment point is abnormal, it is deemed unqualified. VI. Data storage and synchronization.

10. The method and system for identifying surgical sutures based on an industrial camera system according to claim 9, characterized in that, Step three specifically involves: The central processing system performs noise reduction processing on the acquired images A and B through the image processing unit. The noise reduction processing reduces lighting interference through a Gaussian function. The image is then subjected to binarization segmentation. The binarization segmentation method is as follows: (1) Assume the image pixel size is m×n and the gray level L=256. First, calculate the initial threshold T0: ； Where f(x,y) is the gray value of pixel (x,y), and Q is the total number of pixels = m×n; (2) Divide the image into background C0 and foreground C1 according to the initial threshold T0, and calculate the pixel probability of the two types of regions: ； Where, p i =q i / Q, q i Let be the total number of pixels with grayscale value i, and P0 and P1 be the probabilities of C0 and C1, respectively. (3) Calculate the expected gray values ​​of the two types of regions and the overall expected gray value: ； Where μ0 is the expected gray value of C0, μ1 is the expected gray value of C1, and μ is the overall expectation; (4) Maximize the inter-class variance σ B ², Determine the optimal threshold T: ； Anomaly identification was then performed using the following method: Based on the OTSU binarization segmentation results, edge optimization was first performed on the binary image. Then, the image was divided into foreground and background, with the area where the gray value was greater than or equal to the optimal threshold being the foreground of the seam, and the rest being the background of the conveyor belt. Next, the minimum bounding rectangle of the foreground target was obtained through contour detection. The original image was cropped using this rectangle as the boundary to remove irrelevant background areas, resulting in a region of interest (ROI) containing only the seam target. Within the extracted foreground ROI, three types of anomalies were determined: fuzz anomaly determination: if the fuzz length is ≥0.5mm, it is determined to be a fuzz defect, and the corresponding area is marked; thread end anomaly determination: if the thread end diameter is ≥0.1mm, it is determined to be a thread end defect, and the corresponding area is marked; seam diameter deviation anomaly determination: if the diameter deviation is ≥0.05mm, it is determined to be a diameter unevenness defect, and the area with deviation exceeding the limit is marked.

Images