A Machine Vision-Based Pearl Defect Sorting System and Method

Abstract

This invention discloses a pearl defect sorting system and method based on machine vision, belonging to the field of pearl inspection technology. It solves the technical problem that traditional pearl sorting mainly relies on manual inspection, which is inefficient and fails to detect minute defects, leading to unstable pearl quality control. The invention acquires multiple frames of images of the pearls to be sorted; stitches these images together to obtain an image of the pearls to be sorted; identifies the defect areas in the image; calculates the area of ​​the defect areas and analyzes the defect percentage of the sorted pearls; analyzes the luster of the pearls; and grades the pearls according to their defect percentage and luster. This invention demonstrates significant advantages in achieving automated and efficient operation, improving the stability and accuracy of pearl quality assessment.

Description

Technical Field

[0001] This invention belongs to the field of pearl inspection, specifically a pearl defect sorting system and method based on machine vision. Background Technology

[0002] As a natural product, pearls inevitably have various types of imperfections on their surface, such as scratches, spots, and dents. To accurately assess the value and applicability of pearls, it is essential to precisely identify these imperfections. This requires the sorting system to not only distinguish between different types of imperfections but also to quantify the severity of each imperfection in order to perform scientific and reasonable grading.

[0003] Traditional pearl sorting relies primarily on manual inspection, a method that is not only inefficient but also highly dependent on the experience and subjective judgment of the operators. Due to human visual fatigue and individual differences, this manual method struggles to guarantee long-term consistency and accuracy. Furthermore, for large-scale production, manual sorting is costly and fails to meet the demands of rapid market response.

[0004] Therefore, the present invention provides a pearl defect sorting system and method based on machine vision. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a pearl defect sorting system and method based on machine vision, which is used to solve the technical problem that traditional pearl sorting mainly relies on manual inspection, which is inefficient and cannot detect minute defects, resulting in unstable pearl quality control.

[0006] To achieve the above objectives, a first aspect of the present invention provides a pearl defect sorting system based on machine vision, including a processing module, and an image acquisition module and a sorting module connected thereto.

[0007] Image acquisition module: used to acquire multiple frames of images of the pearls to be sorted;

[0008] Processing module: used to stitch together multiple frames of images to obtain the image to be inspected for the pearls to be sorted; and to obtain the defective areas of the image to be inspected.

[0009] The image to be detected is converted from RGB image to LAB color space. The L, A, and B values ​​of each pixel in the converted image are extracted and then calculated using the formula XCZ=α. L value + β A value + γ The defect estimate XCZ of the pixel is calculated by B value; it is determined whether the defect estimate is less than the defect threshold; if yes, the pixel is marked as a defective pixel, and several defective pixels constitute a defective region; otherwise, it is not marked; where α, β and γ are the weight coefficients of L, A and B, respectively.

[0010] The LAB color space of this invention is designed based on the human visual system, where L represents brightness, and A and B represent opposite dimensions of color, such as the green-red axis and the blue-yellow axis. Due to its independent processing characteristics of color information, it makes defect detection more sensitive and accurate. Furthermore, the weights of L, A, and B values ​​can be dynamically adjusted according to the color characteristics and defect type of the pearl to be sorted. This means that for pearls of different colors or pearls undergoing different detection types, the weight parameter settings can be dynamically optimized, thereby improving the accuracy and reliability of defect detection. For example, certain types of defects may be more pronounced in specific color channels; by adjusting the weights of the corresponding channels, the expressiveness of these features can be enhanced.

[0011] The methods for obtaining the defect threshold include:

[0012] Based on the color of the pearl to be sorted, extract several pearl images with the same color as the pearl to be sorted from historical image data; among them, the extracted pearl images are several pearl images with a quality grade of three, and the pearl images are stitched together from multiple consecutive frames of pearl images; the quality grade of pearls is divided into three grades, and the quality grade is grade one > grade two > grade three.

[0013] The level 3 pearl image is divided into several analysis regions. The defect estimate of each pixel in the analysis region is calculated. The median of several defect estimates is extracted as the defect estimate of the analysis region. The mean of the defect estimates of several analysis regions is calculated to obtain the defect estimate of the pixel in the level 3 pearl image, which is marked as the defect threshold.

[0014] This invention selects images of pearls of the same color and quality grade (Level 3) as the pearls to be sorted from historical data as references. This ensures that the set defect threshold provides an effective evaluation benchmark for pearls of specific colors. Different colored pearls may exhibit different defects due to variations in their natural characteristics; this targeted selection better reflects the actual situation. By dividing the Level 3 pearl images into several analysis regions, and using the median, rather than the average, defect estimate as the representative value for each region, outliers, such as extremely high or low defect estimates, can be effectively reduced, improving the robustness of defect assessment. Ensuring the accuracy of the defect threshold is crucial. Then, the defect estimate for each region is calculated, and finally, the average defect estimate for all analysis regions is taken. This method not only considers local details, i.e., the defect situation of individual pixels, but also takes into account overall characteristics, i.e., the defect distribution across the entire pearl surface, thus providing a more comprehensive and detailed defect assessment.

[0015] Calculate the area of ​​the defective region and analyze the proportion of defects in the pearls to be sorted; and analyze the luster of the pearls to be sorted.

[0016] Sorting module: Used to grade the quality of pearls to be sorted based on the percentage of blemishes and their luster.

[0017] Preferably, the step of stitching together multiple frames of images to obtain the detection image of the pearls to be sorted includes:

[0018] An industrial camera captures images of pearls rotating 360° to obtain multiple consecutive frames of images of the pearls to be sorted. These frames are then numbered sequentially according to the direction of rotation. Using panoramic compositing software and the numbered frames, the multiple consecutive frames are stitched together to obtain the image of the pearls to be sorted for inspection.

[0019] This invention uses 360° rotation for imaging, ensuring that every side is captured and obtaining complete appearance information of the pearls to be sorted, avoiding inaccurate evaluations caused by limited perspective. Multiple consecutive frames are sequentially numbered according to the rotation direction and stitched together to generate a high-resolution, richly detailed panoramic image. This is extremely helpful in identifying minute imperfections or features on the pearl's surface, thereby improving sorting accuracy.

[0020] Preferably, the analysis of the percentage of defects in the pearls to be sorted includes:

[0021] Obtain the defect region of the image to be detected, extract the minimum bounding geometry of the defect region, and calculate the area of ​​the defect region according to the area calculation formula of the geometry.

[0022] The ratio between the area of ​​several defective regions and the area of ​​the image to be inspected is calculated to obtain the defect percentage; where several pixels are a unit of length.

[0023] This invention accurately measures the actual size of each defect by precisely locating the defective region in the image to be inspected and extracting its minimum bounding geometry. Compared to traditional visual estimation or coarse measurement methods, this method provides more accurate defect percentage data. The algorithm automatically identifies and calculates the defect percentage, reducing the influence of human factors and ensuring consistent standards in each evaluation. This improved objectivity helps guarantee the consistency and reliability of product quality. The defect percentage data provides strong evidence for quality control and product grading.

[0024] Preferably, the weights of L, A, and B are dynamically adjusted based on the color and blemish type detection of the pearls to be sorted.

[0025] Preferably, the analysis of the luster of the pearls to be sorted includes:

[0026] A luster analysis image of the pearl to be sorted is obtained. Based on the brightness value of each pixel in the luster analysis image, the average brightness of the luster analysis image is calculated. The ratio of the average brightness to a preset brightness threshold is also calculated to obtain the luster of the pearl to be sorted.

[0027] This invention utilizes image processing technology to calculate the average brightness and compare it with a preset brightness threshold. This method provides an objective evaluation standard, avoiding subjective errors caused by human visual judgment. By accurately measuring the brightness value of each pixel and calculating the overall average brightness, the true luster of the pearl can be more accurately reflected. The automated calculation process can quickly process large amounts of data, greatly improving the speed of luster evaluation. Compared to traditional manual inspection methods, this image processing-based method can complete the luster evaluation of multiple pearl samples in a short time, significantly improving work efficiency.

[0028] Preferably, acquiring the luster analysis image of the pearl to be sorted includes:

[0029] The pearls to be sorted are illuminated by a light source, with the light source and the pearls to be sorted positioned on the same horizontal line. An industrial camera is used to capture images of the pearls rotated 360°, and these images are then stitched together to obtain a luster analysis image of the pearls to be sorted.

[0030] This invention places the light source and the pearls to be sorted on the same horizontal line, ensuring uniform light illumination of the pearl surface and reducing shadows and reflection distortion. By rotating the pearl 360° and capturing continuous frame images, the luster of the pearl surface can be captured from all angles without blind spots. This method ensures that the luster characteristics of the pearl can be observed from all angles, thus providing more complete and accurate luster information. This helps to more realistically reflect the actual luster effect of the pearl and improve the accuracy of luster analysis.

[0031] Preferably, the step of calculating the quality index of the pearls to be sorted based on the percentage of defects and the luster of the pearls includes:

[0032] Calculate the normalized values ​​of the percentage of blemishes and luster, and multiply them by the corresponding weighting coefficients to obtain the quality index of the pearl.

[0033] Preferably, a second aspect of the present invention provides a machine vision-based method for sorting pearl defects, comprising the following steps:

[0034] Acquire multiple frames of images of the pearls to be sorted;

[0035] Multiple frames of images are stitched together to obtain the inspection image of the pearls to be sorted; the defect areas of the inspection image are obtained, the area of ​​the defect areas is calculated, and the proportion of defects in the pearls to be sorted is analyzed; and the luster of the pearls to be sorted is analyzed.

[0036] The pearls to be sorted are graded according to the percentage of blemishes and their luster.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] This invention acquires and stitches together multiple frames of images of pearls to be sorted, resulting in a complete and comprehensive image showcasing the pearl's surface features. This ensures detailed observation and analysis of every detail, improving the accuracy and reliability of defect identification. The defect percentage is calculated based on the stitched image, considering not only the number of defects but also quantifying their area relative to the entire pearl surface, providing a more scientific and reasonable defect assessment index. Furthermore, the invention utilizes multiple frames captured under specific settings (e.g., the light source and pearl are on the same horizontal line) for gloss analysis, ensuring that each measurement is performed under identical conditions, thus enhancing the comparability and objectivity of the results. Gloss is assessed by calculating the brightness value of each pixel in the image, capturing minute changes in pearl gloss, which is crucial for accurately evaluating the pearl's gloss characteristics. The entire process, from image acquisition and stitching to defect percentage calculation and gloss analysis, is highly automated, significantly improving work efficiency and reducing labor costs. Quality grading based on these two key parameters—defect percentage and gloss—helps establish a unified and scientific product quality grading system, ensuring product quality consistency and enhancing market competitiveness. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0041] Figure 2 This is a schematic diagram of the defect percentage analysis method of the present invention;

[0042] Figure 3 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0043] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Please see Figure 1The first aspect of the present invention provides a pearl defect sorting system based on machine vision, including a processing module, and an image acquisition module and a sorting module connected thereto;

[0045] The image acquisition module acquires multiple frames of images of the pearls to be sorted;

[0046] The processing module stitches together multiple frames of images to obtain the image to be detected for the pearls to be sorted.

[0047] An industrial camera captures images of pearls rotating 360° to obtain multiple consecutive frames of images of the pearls to be sorted. These frames are then numbered sequentially according to the direction of rotation. Using panoramic compositing software and the numbered frames, the multiple consecutive frames are stitched together to obtain the image of the pearls to be sorted for inspection.

[0048] For example, the image acquisition device of the present invention is configured as follows: an industrial camera, a precision electric rotary table (stepper motor controlled, angular resolution 0.1°; pearl clamp: black matte V-groove to avoid reflection interference, PLC control of rotation synchronized with camera triggering);

[0049] The experimental parameters are set as follows:

[0050] Rotation speed: 10° / second; Shooting interval: camera triggered once every 10° rotation (36 frames in total); Exposure time: 500μs (to avoid motion blur); Image resolution: 2048×2048 pixels;

[0051] Data naming rules:

[0052] The images are numbered according to their rotation angle, such as pearl_001.jpg (0), pearl_002.jpg (1) ... pearl_036.jpg (35);

[0053] Input: 36 consecutive frames (pearl_001.jpg to pearl_036.jpg).

[0054] Steps: Preprocess all images (denoising + equalization); sequentially match adjacent frames (Frame_i) (Frame_i+1); calculate the global transformation (cumulative Homography matrix); project all images onto the same coordinate system to generate a panoramic image.

[0055] Please see Figure 2 The process involves: acquiring the defective regions of the image to be detected; calculating the area of ​​the defective regions; and sorting the pearls by the percentage of defective areas.

[0056] The image to be detected is converted from RGB image to LAB color space. The L, A, and B values ​​of each pixel in the converted image are extracted and then calculated using the formula XCZ=α. L value + β A value + γ The defect estimate XCZ of the pixel is calculated by B value; it is determined whether the defect estimate is less than the defect threshold; if yes, the pixel is marked as a defective pixel, and several defective pixels constitute a defective region; otherwise, it is not marked; where α, β and γ are the weight coefficients of L, A and B, respectively.

[0057] Among them, defect estimation includes defect estimation for different defect types, such as defect estimation for scratch detection and defect estimation for spot detection; therefore, the corresponding defect thresholds include defect thresholds for scratches and defect thresholds for spots.

[0058] This invention further subdivides defect estimation into different types of defect detection, such as scratches and blemishes. This means that each specific type of defect (such as scratches or blemishes) has a corresponding defect estimate obtained by changing the corresponding weighting coefficient. Accordingly, in order to accurately assess whether each type of defect is within an acceptable range, the system also sets specific defect thresholds for different types of defects, such as a scratch defect threshold specifically for judging the severity of scratches, and a blemish defect threshold for assessing the impact of blemishes. This achieves a more detailed and accurate pearl quality assessment, more accurately identifies and distinguishes the specific impact of various defects on pearl quality, thereby supporting more reasonable quality grading decisions.

[0059] The method for obtaining the defect threshold includes the following steps:

[0060] Based on the color of the pearl to be sorted, extract several pearl images with the same color as the pearl to be sorted from historical image data; among them, the extracted pearl images are several pearl images with a quality grade of three, and the pearl images are stitched together from multiple consecutive frames of pearl images; the quality grade of pearls is divided into three grades, and the quality grade is grade one > grade two > grade three.

[0061] The level 3 pearl image is divided into several analysis regions. The defect estimate of each pixel in the analysis region is calculated. The median of several defect estimates is extracted as the defect estimate of the analysis region. The mean of the defect estimates of several analysis regions is calculated to obtain the defect estimate of the pixel in the level 3 pearl image, which is marked as the defect threshold.

[0062] For example, if the pearls to be sorted are white, then select several third-level white pearl images from the historical image database. These images are pearl pictures stitched together from pre-marked consecutive frame images.

[0063] Each level 3 pearl image can be arbitrarily divided into several analysis regions. For example, a 2048×2048 pixel image can be uniformly divided into 50 regions. Small blocks of 50 pixels;

[0064] For all pixels within each analysis region, first convert their RGB values ​​to the LAB color space; calculate the L, A, and B values ​​of each pixel in the LAB color space, and then calculate the defect estimate for each pixel based on certain weights = L. α+A β+B γ; where α, β, and γ are weight coefficients, α+β+γ=1;

[0065] The median of the defect estimates for each pixel within each analysis region is extracted as the defect estimate for that region. The average defect estimate for all analysis regions is then calculated, and this average value is the defect threshold.

[0066] It should be noted that the weights of L, A, and B are dynamically adjusted based on the color and blemish type of the pearls to be sorted.

[0067] For example: 1. If the pearl to be graded is white and the pearl is to be checked for scratches or defects, then give a higher weight to the L value, because the L value represents the brightness information of the image and is particularly sensitive to the detection of brightness. It can detect the scratches of the pearl. Assume the weights are L: 0.6, A: 0.2, B: 0.2.

[0068] When inspecting for scratches and blemishes, it is necessary to increase the weight of A and B values ​​when inspecting for spots and blemishes. This is because A and B correspond to red / green contrast and blue / yellow contrast, respectively, which helps to capture color differences more accurately. At this time, the weights can be adjusted to L:0.2, A:0.4, and B:0.4.

[0069] 2. If the pearl is black, its L value is usually low because black is a low-brightness color. Surface scratches may cause local changes in the L value, but these changes are relatively small and easily overlooked. The color information (A and B values) of black pearls may contain more details, such as surface spots, oxidation marks, or other color abnormalities; color abnormalities (such as brownish-yellow spots or green reflections) are usually the main types of blemishes in black pearls.

[0070] Therefore, for black pearls, the focus should be on color information (A and B values) to detect flaws such as blemishes or oxidation marks. While brightness (L value) still needs to be considered, its weight can be appropriately reduced.

[0071] Obtain the defect region of the image to be detected, extract the minimum bounding geometry of the defect region, and calculate the area of ​​the defect region according to the area calculation formula of the geometry.

[0072] The formula is: Defect percentage = Σ Area of ​​defect region i / Area of ​​the image to be inspected. The defect percentage is calculated at 100%; where several pixels are a unit of length; i is the number of the defect area, and Σ is the summation over i.

[0073] In addition, the luster of the pearls to be sorted was analyzed;

[0074] The pearls to be sorted are illuminated by a light source, with the light source and the pearls to be sorted positioned on the same horizontal line. An industrial camera is used to capture images of the pearls rotated 360°, and these images are then stitched together to obtain a luster analysis image of the pearls to be sorted.

[0075] For example: Equipment preparation, light source: use a ring-shaped LED light source to ensure uniform light distribution and no shadows, suitable for capturing the highlights on the pearl surface; environmental setup: place the light source and the pearls to be sorted on the same horizontal line to ensure that the light directly illuminates the pearl surface and reduces unnecessary reflections or shadow interference; other techniques and steps for image acquisition, processing and analysis are consistent with the defect detection methods discussed earlier.

[0076] A luster analysis image of the pearl to be sorted is obtained. Based on the brightness value of each pixel in the luster analysis image, the average brightness of the luster analysis image is calculated. The ratio of the average brightness to a preset brightness threshold is also calculated to obtain the luster of the pearl to be sorted.

[0077] For example: Convert the gloss analysis image from the RGB color space to the LAB color space. Extract the luminance value (L value) of each pixel in the image. Assuming the image size is 500×500 pixels, there are a total of 500×500=250000 pixels. Calculate the mean luminance value of the gloss analysis image by iterating through all pixels and calculating the average luminance value: Lmean = ∑ i (Li) / M; where Li is the brightness value of the i-th pixel, and M is the total number of pixels in the image (M=250000 in this example). Assume that the mean brightness of the gloss analysis image is calculated to be Lmean=200.

[0078] The ratio of the average brightness to the preset brightness threshold is used as the luster of the pearl: Flaw Degree = Lmean / Ld; where Ld is the brightness threshold, assumed to be 180. Substituting the specific value into the formula: Flaw Degree = 200 / 180≈1.11.

[0079] The sorting module calculates the quality index of the pearls to be sorted based on the percentage of blemishes and the luster, and then grades the pearls according to their quality.

[0080] Using the formula PZS=W1 GZD / GZD max +W2 The quality index PZS of pearls is calculated using (1-XCZ); where GZD represents the percentage of blemishes.max The defect percentage threshold, GZD / GZD max 1 is the normalized value of the defect percentage, XCZ is the gloss, 1-XCZ is the normalized value of the gloss, W1 and W2 are the weighting coefficients, and W1+W2=1.

[0081] The quality index is compared with a preset index threshold range. If the quality index exceeds the preset index threshold range, the quality grade of the pearl to be sorted is Grade 1. If the quality index is within the preset index threshold range, the quality grade of the pearl to be sorted is Grade 2. If the quality index is below the preset index threshold range, the quality grade of the pearl to be sorted is Grade 3. The index threshold range is set by those skilled in the art.

[0082] It should be noted that Grade 1 pearls are of the highest quality, Grade 3 pearls are of the lowest quality, and Grade 2 pearls are of quality between Grade 1 and Grade 2.

[0083] The quality index of the pearl to be sorted in this invention is used to evaluate the quality of the pearl to be sorted. It is mainly related to the proportion of defects and the luster of the pearl to be sorted. Assuming that the weight coefficients of the two are not much different, if the luster is greater and the proportion of defects is smaller, it indicates that the pearl to be sorted has fewer defects and better luster, that is, the quality of the pearl to be sorted is excellent. If the luster is lower and the proportion of defects is larger, it indicates that the pearl to be sorted has more defects and poorer luster, that is, the quality of the pearl to be sorted is poor.

[0084] Please see Figure 3 A second aspect of the present invention provides a machine vision-based method for sorting pearl defects, comprising the following steps:

[0085] Acquire multiple frames of images of the pearls to be sorted;

[0086] Multiple frames of images are stitched together to obtain the inspection image of the pearls to be sorted; the defect areas of the inspection image are obtained, the area of ​​the defect areas is calculated, and the proportion of defects in the pearls to be sorted is analyzed; and the luster of the pearls to be sorted is analyzed.

[0087] The pearls to be sorted are graded according to the percentage of blemishes and their luster.

[0088] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0089] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A pearl defect sorting system based on machine vision, characterized in that, It includes a processing module, as well as an image acquisition module and a sorting module connected to it; Image acquisition module: used to acquire multiple frames of images of the pearls to be sorted; Processing module: used to stitch together multiple frames of images to obtain the image to be inspected for the pearls to be sorted; The percentage of defects in the pearls to be sorted is calculated based on the image to be inspected, including: The image to be detected is converted from RGB to LAB color space. The L, A, and B values ​​of each pixel in the converted image are extracted and then analyzed using the formula... Calculate the defect estimate XCZ of the pixel; determine whether the defect estimate is less than the defect threshold; if yes, mark the pixel as a defective pixel, and several defective pixels constitute a defective region; otherwise, do not mark it; where α, β and γ are the weight coefficients of L, A and B, respectively. The method for obtaining the defect threshold includes: extracting several pearl images with the same color as the pearl to be sorted from historical image data based on the color of the pearl to be sorted; wherein, the extracted pearl images are several pearl images with a quality grade of three, and the pearl images are a stitching of multiple consecutive frames of pearl images; the quality grade of pearls is divided into three grades, and the quality grade is grade one > grade two > grade three. The level 3 pearl image is divided into several analysis regions. The defect estimate of each pixel in the analysis region is calculated. The median of several defect estimates is extracted as the defect estimate of the analysis region. The mean of the defect estimates of several analysis regions is calculated to obtain the defect estimate of the pixel in the level 3 pearl image, which is marked as the defect threshold. Calculate the area of ​​the defective region and analyze the proportion of defects in the pearls to be sorted; and analyze the luster of the pearls to be sorted. The sorting module is used to calculate the normalized values ​​of the percentage of defects and the luster, and multiply them by the corresponding weighting coefficients to obtain the quality index of the pearl; based on the quality index, the pearls to be sorted are graded according to their quality.

2. The pearl defect sorting system based on machine vision according to claim 1, characterized in that, The step of stitching together multiple image frames to obtain the detection image of the pearls to be sorted includes: An industrial camera captures images of pearls rotating 360° to obtain multiple consecutive frames of images of the pearls to be sorted. These frames are then numbered sequentially according to the direction of rotation. Using panoramic compositing software and the numbered frames, the multiple consecutive frames are stitched together to obtain the image of the pearls to be sorted for inspection.

3. The pearl defect sorting system based on machine vision according to claim 1, characterized in that, The analysis of the percentage of defects in the pearls to be sorted includes: Obtain the defect region of the image to be detected, extract the minimum bounding geometry of the defect region, and calculate the area of ​​the defect region according to the area calculation formula of the geometry. The ratio between the area of ​​several defective regions and the area of ​​the image to be inspected is calculated to obtain the defect percentage; where several pixels are a unit of length.

4. The pearl defect sorting system based on machine vision according to claim 1, characterized in that, The weighting coefficients of L, A, and B are dynamically adjusted based on the color and blemish type detection of the pearls to be sorted.

5. The pearl defect sorting system based on machine vision according to claim 1, characterized in that, The analysis of the luster of the pearls to be sorted includes: A luster analysis image of the pearl to be sorted is obtained. Based on the brightness value of each pixel in the luster analysis image, the average brightness of the luster analysis image is calculated. The ratio of the average brightness to a preset brightness threshold is also calculated to obtain the luster of the pearl to be sorted.

6. The pearl defect sorting system based on machine vision according to claim 5, characterized in that, The process of acquiring a luster analysis image of the pearls to be sorted includes: The pearls to be sorted are illuminated by a light source, with the light source and the pearls to be sorted positioned on the same horizontal line. An industrial camera is used to capture images of the pearls rotated 360°, and these images are then stitched together to obtain a luster analysis image of the pearls to be sorted.

7. A pearl defect sorting method based on machine vision, operating based on the pearl defect sorting system based on machine vision according to any one of claims 1-6, characterized in that, Includes the following steps: Acquire multiple frames of images of the pearls to be sorted; Multiple frames of images are stitched together to obtain the inspection image of the pearls to be sorted; the defective areas of the inspection image are obtained, the area of ​​the defective areas is calculated, and the proportion of defects in the pearls to be sorted is analyzed. In addition, the luster of the pearls to be sorted was analyzed; The pearls to be sorted are graded according to the percentage of blemishes and their luster.

Images