A method and system for constructing a production polishing detection model of phosphor copper balls

Abstract

The present invention belongs to the field of industrial modules, and provides a method and system for constructing a polishing detection model for phosphor copper balls, including the steps of collecting optical images and spectral images at the top of a polishing barrel; spatially aligning the optical images and spectral images to determine a first position of the phosphor copper ball in the spectral image, and calculating a corresponding second position of the phosphor copper ball in the optical image; judging the geometric integrity of the phosphor copper ball based on a roundness threshold; calculating a roughness index of the surface of the phosphor copper ball at the second position by using a gray-level co-occurrence matrix analysis method to evaluate whether the polishing is uniform; calculating the uniformity of the surface finish of the phosphor copper ball by using a brightness histogram analysis method to judge whether it meets the polishing standard; when the surface finish of the phosphor copper ball is lower than a preset threshold, adjusting the polishing parameters and continuing the polishing, otherwise stopping the polishing and outputting the detection result.

Description

Technical Field

[0001] The present invention belongs to the field of industrial modules, and particularly relates to a method and system for constructing a polishing detection model for the production of phosphor copper balls. Background Art

[0002] Phosphor copper balls are spherical metal particles widely used in the fields of electronics, welding, and precision manufacturing. Referring to the production line disclosed in the Chinese patent application with the publication number CN118417893A, its production process generally includes multiple links such as raw material processing, spheroidization forming, polishing treatment, and quality inspection. Polishing, as a key step to improve the surface finish of phosphor copper balls, reduce the oxide layer, and improve the electrical conductivity, directly affects the final quality of the product. However, in the traditional phosphor copper ball polishing process, there are no effective polishing detection means, resulting in low polishing efficiency, unstable quality, and a high degree of manual dependence, making it difficult to meet the requirements of modern industry for automated and efficient production.

[0003] In the existing phosphor copper ball production process, a long polishing method with a fixed time is usually adopted, that is, without real-time detection of the polishing state of the phosphor copper balls, a relatively long polishing time is set to ensure that all phosphor copper balls can reach the qualified surface finish. However, the fixed-time polishing method often sets a relatively long polishing time to achieve a good polishing effect, thereby reducing the overall production efficiency and increasing unnecessary energy consumption and material consumption.

[0004] Manual inspection is still the quality control method adopted by some current phosphor copper ball production enterprises, but it is difficult to meet the requirements of large-scale continuous production.

[0005] In addition, the existing machine vision detection technology faces major challenges during the polishing process of phosphor copper balls. Since the phosphor copper balls are in a high-density stacked state in the polishing barrel, and chemical substances such as abrasive sand, brightening agent, and automation agent are added during the polishing process, these substances will form a covering layer or liquid film, seriously affecting the imaging quality of the optical detection system for the surface of the phosphor copper balls. For example, during the polishing process, the brightening agent and automation agent will form a liquid film on the surface of the phosphor copper balls, generating strong specular reflection during optical imaging, making it difficult for traditional industrial cameras to clearly capture the true surface state of the phosphor copper balls, resulting in detection misjudgment or inability to identify effective targets. The abrasive sand particles form a randomly distributed covering layer on the surface of the phosphor copper balls, making the proportion of the phosphor copper balls in the picture too small to be detected. The internal lighting environment of the polishing barrel is complex, and the scattering and reflection characteristics of the light source will cause large brightness unevenness in the captured images, making it difficult for the image detection algorithm based on gray-scale features to distinguish the surface of the phosphor copper balls from the background and affecting the detection accuracy. Summary of the Invention

[0006] To solve the problems in the prior art, the present invention provides a method for constructing a polishing detection model for phosphor copper balls, comprising the following steps:

[0007] Step S10, data acquisition: Use an industrial camera to obtain the optical image of the phosphor copper balls at the top of the polishing barrel, and use a short-wave infrared hyperspectral camera to obtain the spectral image at the top of the polishing barrel;

[0008] Step S20, spatially align the optical image and the spectral image to analyze them in the same coordinate system; Determine the first position of the phosphor copper balls in the spectral image through the target detection algorithm, and calculate the corresponding second position of the phosphor copper balls in the optical image based on the coordinate mapping model of the optical image to ensure the consistency of the same phosphor copper balls in the two data channels;

[0009] Step S30, use the ellipse fitting algorithm to calculate the roundness of the phosphor copper balls at the second position, and judge the geometric integrity of the phosphor copper balls based on the roundness threshold. When the roundness of the phosphor copper balls does not meet the standard, adjust the polishing parameters and continue grinding;

[0010] Step S40, use the gray-level co-occurrence matrix analysis method to calculate the surface roughness index of the phosphor copper balls at the second position to evaluate whether the polishing is uniform; Use the brightness histogram analysis method to calculate the uniformity of the surface finish of the phosphor copper balls and judge whether it meets the polishing standard; When the surface finish of the phosphor copper balls is lower than the preset threshold, adjust the polishing parameters and continue polishing, otherwise stop polishing and output the detection result.

[0011] The present invention also provides a system for constructing a polishing detection model for phosphor copper balls, comprising the following modules:

[0012] The data acquisition module is used to obtain the optical image of the phosphor copper balls at the top of the polishing barrel by using an industrial camera, and obtain the spectral image at the top of the polishing barrel by using a short-wave infrared hyperspectral camera;

[0013] The calibration module is used to spatially align the optical image and the spectral image to analyze them in the same coordinate system; Determine the first position of the phosphor copper balls in the spectral image through the target detection algorithm, and calculate the corresponding second position of the phosphor copper balls in the optical image based on the coordinate mapping model of the optical image to ensure the consistency of the same phosphor copper balls in the two data channels;

[0014] The first judgment module is used to use the ellipse fitting algorithm to calculate the roundness of the phosphor copper balls at the second position, and judge the geometric integrity of the phosphor copper balls based on the roundness threshold. When the roundness of the phosphor copper balls does not meet the standard, adjust the polishing parameters and continue grinding;

[0015] The second judgment module is used to calculate the surface roughness index of the phosphor copper ball at the second position by using the gray-level co-occurrence matrix analysis method to evaluate whether the polishing is uniform; calculate the smoothness uniformity of the phosphor copper ball by using the brightness histogram analysis method to judge whether it meets the polishing standard; when the smoothness of the phosphor copper ball is lower than the preset threshold, adjust the polishing parameters and continue polishing, otherwise stop polishing and output the detection result.

[0016] By constructing a polishing detection model for phosphor copper balls based on the fusion of spectral images and optical images, the present invention overcomes the problems of low detection accuracy, uncontrollable polishing time, and low efficiency of manual detection in the existing polishing process, and has the following beneficial effects:

[0017] Identify the position of the phosphor copper ball through the spectral image and map it to the optical image to achieve precise positioning, avoid detection errors caused by the coverage of abrasive, brightening agent, and automation agent, and improve the detection stability and robustness. By calculating the roundness, roughness, and smoothness of the phosphor copper ball in real time, dynamically adjust the polishing parameters, avoid unnecessary long-time polishing, improve the automation level of the production line, and reduce energy and material waste. Use computer vision and intelligent algorithms to replace manual detection, reduce human misjudgment, improve the stability and consistency of detection standards, ensure that the polishing quality of phosphor copper balls meets high-precision requirements, and at the same time reduce the labor cost input of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0019] Figure 1 is the phosphor copper ball production line targeted by the present invention;

[0020] Figure 2 is the polishing device targeted by the present invention;

[0021] Figure 3 is the method flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] Next, with reference to the drawings and specific embodiments, a preferred description of the invention will be made.

[0023] As Figure 1 shown, this embodiment describes a further optimization model of a phosphor copper ball production line disclosed in a Chinese patent application with the publication number CN118417893A.

[0024] The technological process for producing copper balls is as follows: feeding tray → ball heading → conveying → collecting → polishing → cleaning → drying → weighing → packaging → inkjet coding → palletizing.

[0025] Among them, the automatic ball heading machine:

[0026] The ball heading process is: feeding tray (disc coiled material) → straightening machine → cutting → ball heading;

[0027] Among them, the automatic polishing production line:

[0028] The polishing process is: balls out from the ball heading machine → lifting conveyor → aggregate bucket → diversion trough body → grinding machines (2 sets) → spray pipe → discharging.

[0029] Among them, the automatic cleaning machine line:

[0030] The cleaning process is: feeding → high-pressure spraying → water blowing by fan → drying → feeding by conveyor belt → discharging.

[0031] Among them, the automatic packaging production line:

[0032] Discharging from the cleaning machine → lifting conveyor → automatic weighing → conveying → case sealing → conveying → inkjet coding → conveying → strapping → conveying → stacking.

[0033] As Figure 2 shown is a diagram of the polishing mechanism. Among them, phosphor bronze balls are polished inside the polishing barrel. During the polishing process, abrasive sand, brightening agent, automation agent, etc. will be used.

[0034] In one embodiment, referring to Figure 3 , the present invention provides a method for constructing a polishing detection model for phosphor bronze balls. By collecting the optical images and spectral data of phosphor bronze balls in the polishing barrel, data fusion and analysis are carried out to construct a detection model capable of accurately identifying the polishing state of phosphor bronze balls, and the polishing parameters are adjusted in combination with intelligent optimization algorithms to ensure the polishing uniformity and quality consistency of the surface of phosphor bronze balls and improve the detection accuracy and production efficiency.

[0035] In the present invention, phosphor bronze balls are spherical metal particles made of phosphor bronze material. Its production process includes multiple links such as raw material processing, forming, polishing, and surface quality detection. Polishing refers to removing the oxide layer, burrs, and microscopic unevenness on the surface of phosphor bronze balls through mechanical grinding, chemical treatment, or other means to improve its smoothness and surface uniformity to meet specific usage requirements.

[0036] Polishing detection refers to, during the polishing process of phosphor bronze balls, real-time monitoring of the surface morphology, smoothness, roundness, and defect conditions of phosphor bronze balls through optical imaging, spectral analysis, or other detection technologies, and analyzing the detection results to determine whether the phosphor bronze balls meet the polishing standards.

[0037] Based on the data collected during the polishing process of phosphor bronze balls, this invention utilizes technologies such as computer vision, deep learning, spectral analysis, and statistical analysis to establish a mathematical model, optimize the detection algorithm, and combine a feedback control strategy to achieve dynamic evaluation and intelligent regulation of the polishing quality of phosphor bronze balls. The specific steps are as follows:

[0038] Step S10: Data acquisition. An industrial camera is used to obtain the optical image of the phosphor bronze ball at the top of the polishing barrel, and a short-wave infrared hyperspectral camera is used to obtain the spectral image at the top of the polishing barrel.

[0039] Optical image acquisition refers to using an industrial camera to obtain the visible light image of the phosphor bronze ball at the top of the polishing barrel to record the geometric features, contour morphology, and surface texture information of the phosphor bronze ball. An industrial camera is a high-resolution imaging device for precise measurement and detection. Compared with ordinary cameras, it has higher imaging quality, faster frame rate, and stronger environmental adaptability. The imaging area of the industrial camera is set at the top of the polishing barrel, enabling it to capture the phosphor bronze balls in the upper layer during the polishing process. Combining with external light source compensation technology ensures uniform illumination and reduces environmental light interference to improve imaging quality.

[0040] Spectral image acquisition refers to using a short-wave infrared hyperspectral camera to obtain the spectral image of the phosphor bronze ball at the top of the polishing barrel to capture the reflectivity information of the phosphor bronze ball at different wavelengths, thereby analyzing its surface state. A short-wave infrared hyperspectral camera is a detection device that can simultaneously image in multiple wavelength channels. Its principle is based on spectral decomposition technology, which collects the reflected light of the target object in multiple bands to obtain richer material characteristic information than visible light imaging. During the polishing process of phosphor bronze balls, due to the presence of brighteners and activators, the surface reflection characteristics may change. Short-wave infrared imaging can effectively distinguish the smooth surface from the areas covered by brighteners and activators.

[0041] Step S20: Align the optical image and the spectral image spatially so that they can be analyzed in the same coordinate system. Determine the first position of the phosphor bronze ball in the spectral image through the target detection algorithm, and calculate the corresponding second position of the phosphor bronze ball in the optical image based on the coordinate mapping model of the optical image to ensure the consistency of the same phosphor bronze ball in the two data channels.

[0042] During the polishing process of phosphor bronze balls, due to the use of abrasive sand, brighteners, and activators, most of the surface of the acquired images is covered by abrasive sand and the liquid film generated by brighteners and activators. The liquid film formed by the brightener produces strong specular reflection, making it difficult for ordinary image processing algorithms to distinguish whether the target is a phosphor bronze ball or a liquid film.

[0043] Due to the above factors, it is difficult to directly identify the position of the phosphor bronze ball from ordinary optical images, making traditional algorithms based on edge detection or target detection perform unstably in complex polishing environments.

[0044] The spectral characteristics of phosphor bronze balls are significantly different from those of abrasive sand, brighteners, and activators. In the short-wave infrared (SWIR) band, the spectral reflectivity of phosphor bronze material differs greatly from that of abrasive sand, brighteners, and activators, enabling the spectral camera to clearly distinguish phosphor bronze balls from the background.

[0045] Abrasive sand usually exhibits low infrared reflectivity. Its particle morphology leads to enhanced scattering, forming high-contrast regions in the spectral image, which is significantly different from the smooth surface of phosphor bronze balls. Brighteners may cause high reflectivity in the visible light range but usually show relatively uniform low-reflectivity regions in the short-wave infrared band, so they do not interfere with the positioning of phosphor bronze balls in the spectral image. Due to different chemical compositions, the absorption characteristics of the deposition layer of the activator are significantly different from those of phosphor bronze material at specific infrared wavelengths, enabling the spectral camera to identify the position of phosphor bronze balls through multi-band imaging.

[0046] However, due to the low gray-scale contrast of spectral data, it is difficult to accurately segment the boundary contour of phosphor bronze balls in the spectral image. Roundness detection requires high-precision boundary data for ellipse fitting and shape analysis, so the spectral image is not suitable for roundness detection. At the same time, since spectral images are analyzed based on the spectral characteristics of material components, the reflectivity of light at different wavelengths may be affected by the residues of abrasive sand, brighteners, and activators, resulting in misjudgment of surface finish evaluation.

[0047] Based on the respective advantages of optical images and spectral images, the following steps can be used to achieve the auxiliary positioning of spectral images to optical images:

[0048] First, detect the position of phosphor bronze balls in the spectral image. Use an object detection algorithm based on spectral features to accurately identify the position of phosphor bronze balls in the spectral image and determine its first position, that is, the coordinate points of the phosphor bronze balls in the spectral image. Through a deep learning object detection model (such as YOLO or Faster R-CNN), extract the unique spectral features of phosphor bronze balls from multi-band data and establish a classification model, enabling the detection system to exclude the interference of abrasive sand, brighteners, and activators.

[0049] Based on the coordinate mapping model, map the first position of the spectral image to the second position of the optical image. Since the imaging angles and resolutions of the optical camera and the spectral camera are different, the coordinate positions of the same phosphor bronze ball in the two images may deviate. Therefore, a spectral-optical coordinate mapping model needs to be established.

[0050] In the present invention, the relative positions of the optical camera and the spectral camera are fixed, that is, the two cameras will not undergo physical displacement during the entire polishing detection process of the phosphor bronze ball. Therefore, a fixed spatial mapping relationship can be established based on the camera coordinate system, enabling the optical image and the spectral image to be analyzed in the same coordinate system, thereby improving the stability and consistency of the detection.

[0051] Since the positions of the optical camera and the spectral camera are fixed, a one-time calibration can be performed during system installation to establish the coordinate transformation relationship between the two cameras, ensuring that the images collected by both can be mapped to a unified coordinate system. The specific construction method of the spectral-optical coordinate mapping model is as follows:

[0052] First, perform the calibration of the camera internal parameters. Using a checkerboard or calibration board, calibrate the internal parameter matrices of the optical camera and the spectral camera respectively, including the focal length, principal point coordinates, and lens distortion coefficients.

[0053] Using a checkerboard or calibration board, calibrate the optical camera and the spectral camera respectively to obtain the internal parameter matrix of the camera and the lens distortion coefficient.

[0054] The internal parameter matrix is defined as follows:

[0055]

[0056] Where:

[0057] represents the focal lengths in the horizontal and vertical directions (unit: pixel);

[0058] represents the principal point coordinates (optical center).

[0059] The lens distortion parameters are

[0060]

[0061] where k 1 , k 2 , k 3 , k 4 are the radial distortion coefficients;

[0062] are the tangential distortion coefficients.

[0063] During the calibration process, multiple checkerboard images at different angles can be collected, and then the cv2.calibrateCamera() of OpenCV can be used to calculate the internal parameter matrix of the camera and the distortion coefficient .

[0064] Then, perform the extrinsic calibration of the camera to determine the relative position relationship (i.e., the rotation matrix R and the translation vector T) between the optical camera and the spectral camera, which is used to describe the coordinate transformation between the two cameras.

[0065] Among them, the relationship between the rotation matrix and the translation vector is as follows:

[0066]

[0067] Among them:

[0068] R represents the rotation matrix; T represents the translation vector; represents the point coordinates in the optical image, represents the point coordinates in the spectral image.

[0069] During the specific calibration, use a calibration board to take multiple groups of images in the overlapping field of view of the optical camera and the spectral camera, and then calculate the corresponding relationship of the checkerboard corner points between the cameras. Use the PnP algorithm to solve the extrinsic parameters and calculate the extrinsic parameter matrix using cv2.solvePnP():

[0070] Calculate the homography matrix between the cameras. Since the cameras are fixed, the projection transformation matrix of the two can be obtained through one-time calibration and stored for subsequent image conversion; in the formula, object points represent the three-dimensional coordinates of the points calibrated on the calibration board; image points represent the pixel coordinates corresponding to the projection of object points in the camera image.

[0071] Then, the image alignment can be directly performed using the camera calibration results without dynamic registration during each detection. Using the calculated rotation matrix and translation vector, convert the pixel points in the spectral image to the coordinate system of the optical image to ensure the spatial position matching of the two.

[0072] First, calculate the 3D coordinates of each pixel point in the spectral image

[0073] Assume the point in the spectral image is in the pixel coordinate system, then there is:

[0074]

[0075] Among them:

[0076] is the inverse matrix of the intrinsic parameter matrix of the spectral camera;

[0077] The 3D point in the spectral camera coordinate system;

[0078] Project the points in the spectral image onto the optical camera coordinate system. When transforming to the optical camera coordinate system, we have:

[0079]

[0080] Then re-project it onto the optical image:

[0081]

[0082] Where, is the optical image obtained by re-projection;

[0083] Obtain the coordinates of the transformed optical image.

[0084] Then, through a homography transformation, use the projection transformation matrix to calculate the corresponding positions of each pixel point in the spectral image in the optical image and perform remapping.

[0085] Specifically, first calculate the homography matrix H:

[0086]

[0087] Where:

[0088] is the internal parameter of the optical camera;

[0089] is the plane normal vector;

[0090] is the distance from the plane to the camera

[0091] is the transformation matrix.

[0092] Then perform the projection transformation

[0093]

[0094] Where:

[0095] are the pixel coordinates of the spectral image;

[0096] are the mapped coordinates in the optical image;

[0097] represents the perspective scaling factor.

[0098] Then use cv2.warpPerspective() to perform a perspective transformation to re-project the spectral image onto the optical image coordinate system.

[0099] By using the aligned image coordinates, the phosphor bronze ball information in the spectral image is marked on the optical image to assist subsequent geometric detection.

[0100] Through the second position provided by the spectral image, region cropping is performed in the optical image, and fine-grained detection is only carried out for the positions where phosphor bronze balls may exist, avoiding the computational overhead caused by global search.

[0101] In this step, the spectral image is used to assist in positioning the optical image, overcoming the detection difficulties caused by the coverage of abrasive sand, brightening agent, and automation agent, and improving the accuracy of phosphor bronze ball polishing detection. By constructing a spectral-optical coordinate mapping model, the consistency of the same phosphor bronze ball in the two data channels is ensured, making the subsequent polishing state analysis more accurate. This method can effectively improve the stability of the phosphor bronze ball polishing detection system and provide important data support for optimizing the polishing process and improving product quality.

[0102] In step S30, the roundness of the phosphor bronze ball at the second position is calculated using the ellipse fitting algorithm, and the geometric integrity of the phosphor bronze ball is judged based on the roundness threshold. When the roundness of the phosphor bronze ball does not meet the standard, the polishing parameters are adjusted and grinding continues.

[0103] During the polishing process, the roundness of the phosphor bronze ball gradually increases. When its roundness approaches the standard value, the surface microstructure tends to be smooth, causing the abrasive sand, brightening agent, and automation agent attached to the surface of the phosphor bronze ball to be difficult to continue adhering and instead naturally slide off or be thrown out by the high-speed rotating polishing barrel.

[0104] Due to the low surface tension of the polished phosphor bronze ball, the residual grinding medium cannot adhere effectively, making the phosphor bronze ball in the optical image present a standard circular area, that is, the boundary of the phosphor bronze ball becomes clearly visible and forms an obvious contrast with the background.

[0105] This phenomenon can be used to further improve the detection accuracy, that is, when the contour of the phosphor bronze ball in the image presents a complete and unbroken standard circular area, it can be judged that its polishing is close to the completion state.

[0106] Since the specific position of the phosphor bronze ball has been located in the previous step, the roundness of the phosphor bronze ball is calculated based on the ellipse fitting algorithm. The ellipse fitting algorithm refers to calculating the ellipse parameters and evaluating the fitting error on the detected contour point set of the phosphor bronze ball through the least squares method or the ellipse least squares fitting method to determine whether the geometric shape of the phosphor bronze ball is close to an ideal circle.

[0107] Specifically, first, the contour points of the phosphor bronze ball are extracted through an edge detection method (such as the Canny algorithm), and then the major axis, minor axis, center point coordinates, and rotation angle of the phosphor bronze ball are calculated based on the ellipse fitting algorithm, and whether the shape of the phosphor bronze ball meets the standard is determined through the roundness calculation formula.

[0108] The roundness threshold refers to a preset value used to evaluate whether the phosphor bronze ball meets the standard shape. Its value is usually set based on experimental data and corresponds to the quality requirements of the target product.

[0109] When the calculated roundness is lower than the preset threshold, it indicates that there are still minor deformations on the surface of the phosphor bronze ball, which may be caused by uneven polishing, uneven distribution of grinding media, or insufficient polishing time.

[0110] At this time, the system automatically adjusts the polishing parameters, such as increasing the grinding time, increasing the rotation speed of the polishing barrel, or optimizing the ratio of grinding sand, brightening agent, and activator, to further improve the geometric shape of the phosphor bronze ball.

[0111] In this step, the roundness is calculated through the elliptical fitting algorithm, which can monitor the morphological changes of the phosphor bronze ball in real time during the polishing process and dynamically adjust the polishing parameters according to the detection results to improve the polishing accuracy. Due to the natural sliding phenomenon of the grinding media on the surface of the phosphor bronze ball, when the phosphor bronze ball reaches the standard roundness, the system can automatically identify and stop polishing, avoiding unnecessary resource consumption and improving production efficiency. At the same time, through the standardized analysis of the phosphor bronze ball area in the optical image, the subsequent surface finish detection and surface defect identification can be ensured to be more accurate.

[0112] In step S40, the gray-level co-occurrence matrix analysis method is used to calculate the surface roughness index of the phosphor bronze ball at the second position to evaluate whether the polishing is uniform; the brightness histogram analysis method is used to calculate the surface finish uniformity of the phosphor bronze ball to determine whether it meets the polishing standard; when the surface finish of the phosphor bronze ball is lower than the preset threshold, the polishing parameters are adjusted and polishing continues, otherwise polishing stops and the detection results are output.

[0113] The gray-level co-occurrence matrix analysis method refers to calculating the spatial relationship between pixel points of the image gray level, extracting the texture features on the surface of the phosphor bronze ball, and using this to measure whether the polishing is uniform.

[0114] Specifically, first, the optical image of the phosphor bronze ball is subjected to gray normalization processing, a fixed pixel window is set, and the gray-level co-occurrence matrix between adjacent pixels is calculated in different directions (0°, 45°, 90°, 135°), and the surface roughness is quantified by extracting features such as contrast, energy, homogeneity, and entropy.

[0115] Specifically, since the optical image may contain pixel values in different brightness ranges, it is necessary to perform gray normalization on the image to ensure consistent image contrast under different lighting conditions.

[0116] Convert the optical image into a grayscale image

[0117]

[0118] Wherein:

[0119] are the red, green, and blue channels of the image

[0120] is the converted grayscale image.

[0121] Then perform grayscale normalization

[0122]

[0123] Wherein:

[0124] and are respectively the minimum and maximum grayscale values of the original image;

[0125] and is the grayscale range after normalization (0 - 255)

[0126] represents the grayscale image after normalization.

[0127] Set the window size to (such as 5×5 or 7×7), slide the window over the entire image, and calculate the gray-level co-occurrence matrix for each window.

[0128] Gray-level co-occurrence matrix is determined by the following method:

[0129]

[0130] Wherein, is the gray level (such as 0 - 255);

[0131] are the pixel coordinates of the image;

[0132] represents the relative position (direction) between pixels, wherein:

[0133] 0° horizontal direction

[0134] 45° diagonal direction

[0135] 90° vertical direction

[0136] 135° diagonal direction

[0137] The gray-level co-occurrence matrix is a dimensional matrix ( is the number of gray levels, such as 256), where each element represents the frequency of the co-occurrence relationship between gray levels to .

[0138] Contrast measures the intensity of the gray-level difference and reflects whether there are rough or uneven areas on the surface:

[0139]

[0140] Energy measures the degree of uniformity. The higher the energy value, the more regular the surface texture:

[0141]

[0142] Homogeneity measures the smoothness of the distribution of adjacent gray levels. A surface with high homogeneity is smoother:

[0143]

[0144] Entropy measures the amount of information and reflects the randomness and complexity of the texture:

[0145]

[0146] Among them, the contrast value reflects the surface height change. The larger the value, the higher the surface roughness; the energy and homogeneity indicators reflect the consistency of the surface texture. The higher the value, the more uniform the polishing.

[0147] When the calculated roughness is higher than the preset threshold, it indicates that there are still minor wear unevenness or grinding residues on the surface of the phosphor bronze ball. At this time, adjust the polishing parameters to optimize the grinding process.

[0148] The brightness histogram analysis method refers to judging the uniformity of the surface finish by statistically analyzing the distribution of the brightness values of different pixels on the surface of the phosphor bronze ball.

[0149] Specifically, first, the brightness distribution of the phosphor bronze ball area in the optical image is statistically analyzed, and the standard deviation, kurtosis, and skewness of the brightness histogram are calculated to evaluate whether the surface finish is uniform.

[0150] Among them, the standard deviation of the brightness histogram reflects the degree of dispersion of the brightness distribution. When the value is low, it indicates that the surface finish is uniform; the kurtosis represents the degree of concentration of the image brightness. The higher the value, the better the polishing uniformity; the skewness is used to judge the symmetry of the brightness distribution. If the skewness is close to zero, it means that the surface finish is uniform.

[0151] If the surface finish is lower than the preset threshold, it indicates that there are still local rough areas or grinding residues on the surface of the phosphor copper ball, and at this time, polishing needs to be continued.

[0152] When the test results show that the surface roughness of the phosphor copper ball is high or the surface finish is low, the system automatically adjusts the polishing parameters, including:

[0153] Adjust the fineness of the grinding sand. If the surface roughness is relatively high, increase the proportion of fine grinding sand to improve the surface smoothness.

[0154] Optimize the ratio of brightener to activator. If the surface finish is uneven, adjust the concentration of the brightener to improve the uniformity of surface light reflection.

[0155] Prolong the polishing time. If the surface finish does not meet the standard, appropriately increase the polishing time to ensure that the surfaces of all phosphor copper balls uniformly meet the polishing requirements.

[0156] Optimize the rotation speed of the polishing barrel. If the surface finish is still low, adjust the rotation mode of the polishing barrel to make the phosphor copper balls tumble more uniformly in the barrel and improve the grinding uniformity.

[0157] When the test results show that the roughness of the phosphor copper ball is lower than the preset threshold and the uniformity of the surface finish is higher than the set standard, the system determines that the polishing of the phosphor copper ball is completed and records its test results.

[0158] In this step, the surface roughness of the phosphor copper ball is calculated through the gray-level co-occurrence matrix, and the uniformity of the surface finish is analyzed in combination with the luminance histogram to construct a high-precision polishing detection system. And the polishing parameters are adjusted through intelligent feedback to ensure the uniform surface quality of the phosphor copper balls, optimize the polishing production process, and improve the overall production efficiency and quality stability.

[0159] In another embodiment, the present invention also provides a system for constructing a polishing detection model for phosphor copper balls, including:

[0160] A data acquisition module for obtaining an optical image of the phosphor copper ball at the top of the polishing barrel using an industrial camera and obtaining a spectral image at the top of the polishing barrel using a short-wave infrared hyperspectral camera;

[0161] A calibration module for spatially aligning the optical image and the spectral image so that they can be analyzed in the same coordinate system; determining the first position of the phosphor copper ball in the spectral image through a target detection algorithm, and calculating the corresponding second position of the phosphor copper ball in the optical image based on the coordinate mapping model of the optical image to ensure the consistency of the same phosphor copper ball in the two data channels;

[0162] The first judgment module is used to calculate the roundness of the phosphor copper ball at the second position by using the elliptical fitting algorithm, and judge the geometric integrity of the phosphor copper ball based on the roundness threshold. When the roundness of the phosphor copper ball does not meet the standard, adjust the polishing parameters and continue grinding;

[0163] The second judgment module is used to calculate the surface roughness index of the phosphor copper ball at the second position by using the gray level co-occurrence matrix analysis method to evaluate whether the polishing is uniform; use the brightness histogram analysis method to calculate the smoothness uniformity of the phosphor copper ball and judge whether it meets the polishing standard; when the smoothness of the phosphor copper ball is lower than the preset threshold, adjust the polishing parameters and continue polishing, otherwise stop polishing and output the detection result.

[0164] In a further embodiment, the spatial alignment of the optical image and the spectral image is achieved through the following modules:

[0165] The internal parameter calibration module is used to perform camera internal parameter calibration. Using a checkerboard or calibration board, calibrate the internal parameter matrices of the optical camera and the spectral camera respectively, including the focal length, the principal point coordinates and the lens distortion coefficients;

[0166] The external parameter calibration module is used to perform camera external parameter calibration, determine the relative position relationship between the optical camera and the spectral camera, that is, the rotation matrix and the translation vector, which are used to describe the coordinate transformation between the two cameras, and calculate the homography matrix between the cameras;

[0167] The spatial position matching module is used to use the calculated rotation matrix and translation vector to convert the pixel points in the spectral image into the coordinate system of the optical image to ensure the spatial position matching of the two;

[0168] The remapping module is used to further perform homography transformation, calculate the corresponding position of each pixel point in the spectral image in the optical image by using the projection transformation matrix, and perform remapping.

[0169] In a further embodiment, the calculation of the surface roughness index of the phosphor copper ball by using the gray level co-occurrence matrix analysis method is achieved through the following modules:

[0170] The image preprocessing module is used to perform gray normalization processing on the optical image of the phosphor copper ball, set a fixed pixel window, calculate the gray level co-occurrence matrix between adjacent pixels in different directions, and quantify the surface roughness by extracting features such as contrast, energy, uniformity, and entropy; the grinding judgment module is used to when the calculated roughness is higher than the preset threshold, indicating that there are still slight wear unevenness or grinding residues on the surface of the phosphor copper ball, at this time adjust the polishing parameters and continue grinding.

[0171] In a further embodiment, the calculation of the smoothness uniformity of the phosphor copper ball by using the brightness histogram analysis method is implemented by using the following module:

[0172] A brightness distribution evaluation module is used to statistically analyze the brightness distribution of the phosphor bronze ball area in an optical image, and calculate the standard deviation, kurtosis, and skewness of the brightness histogram to evaluate whether the surface finish is uniform.

[0173] A brightness histogram evaluation module: The standard deviation of the brightness histogram reflects the degree of dispersion of the brightness distribution. The lower the value, the more uniform the surface finish. The kurtosis represents the degree of concentration of the image brightness. The higher the value, the better the polishing uniformity. The skewness is used to judge the symmetry of the brightness distribution. The lower the value, the more uniform the surface finish.

[0174] A finish evaluation module: If the finish is lower than the preset threshold, it indicates that there are still local rough areas or grinding residues on the surface of the phosphor bronze ball, and at this time, polishing needs to be continued.

[0175] A polishing judgment module is used to adjust the polishing parameters and continue polishing when the detection results show that the surface roughness of the phosphor bronze ball is high or the finish is low.

[0176] It should be noted that the explanatory description of the foregoing method embodiments for constructing the polishing detection model of the phosphor bronze ball also applies to the device of the embodiments of the present application, and will not be elaborated here.

[0177] Those of ordinary skill in the art can realize that the units and algorithm steps described in the embodiments disclosed herein can be implemented by a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0178] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated here.

[0179] In several embodiments provided by the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (hereinafter referred to as ROM), random access memory (hereinafter referred to as RAM), magnetic disks, or optical discs that can store program codes.

[0180] As described above, the foregoing is only the specific implementation manner of the present application. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. The protection scope of the present application shall be subject to the protection scope of the claims. For the part of the module structure that is not specifically defined in the present invention, the content recorded in the prior art shall prevail. The prior art mentioned in the foregoing background art part and the specific embodiment part of the present invention can be used as a part of the present invention to understand the meaning of some technical features or parameters.

Claims

1. A method for constructing a production polishing detection model for phosphor bronze balls, characterized in that: The method comprises the following steps: Step S10, data acquisition, using an industrial camera to obtain an optical image of the phosphor bronze ball on the top of the polishing barrel, and using a short-wave infrared hyperspectral camera to obtain a spectral image of the top of the polishing barrel; Step S20, spatially aligning the optical image and the spectral image so that the two are analyzed in the same coordinate system; determining the first position of the phosphor bronze ball in the spectral image by a target detection algorithm, and calculating the corresponding second position of the phosphor bronze ball in the optical image based on the coordinate mapping model of the optical image, to ensure the consistency of the same phosphor bronze ball in the two data channels; Step S30, using an ellipse fitting algorithm to calculate the roundness of the phosphor bronze ball at the second position, and judging the geometric integrity of the phosphor bronze ball based on a roundness threshold, and when the roundness of the phosphor bronze ball does not meet the standard, adjusting the polishing parameters and continuing the grinding; Step S40, using a gray level co-occurrence matrix analysis method to calculate the roughness index of the surface of the phosphor bronze ball at the second position to evaluate whether the polishing is uniform; using a brightness histogram analysis method to calculate the smoothness uniformity of the phosphor bronze ball to determine whether it meets the polishing standard; when the smoothness of the phosphor bronze ball is lower than a preset threshold, adjusting the polishing parameters and continuing the polishing, otherwise stopping the polishing and outputting the test result; The positions of the industrial camera and the short-wave infrared hyperspectral camera are relatively fixed; The spatial alignment of the optical image and the spectral image comprises: Perform camera internal parameter calibration, using a chessboard or calibration plate to calibrate the internal parameter matrices of the optical camera and spectral camera, including focal length, principal point coordinates, and lens distortion coefficients; Perform camera extrinsic calibration to determine the relative position relationship between the optical camera and the spectral camera, that is, the rotation matrix and translation vector, which are used to describe the coordinate transformation of the two cameras and calculate the homography matrix between the cameras; The calculated rotation matrix and translation vector are used to transform the pixels in the spectral image into the coordinate system of the optical image to ensure that the spatial positions of the two match. Then, through homography transformation, the corresponding position of each pixel in the spectral image in the optical image is calculated using the projection transformation matrix and remapped.

2. The method for constructing a phosphor bronze ball production polishing detection model according to claim 1, characterized in that: The gray level co-occurrence matrix analysis method is used to calculate the roughness index of the phosphor bronze ball surface, including: The optical image of the phosphor bronze ball is grayscale normalized, a fixed pixel window is set, the grayscale co-occurrence matrix between adjacent pixels is calculated in different directions, and the surface roughness is quantified by extracting contrast, energy, uniformity, and entropy features; The contrast value reflects the change in surface height. The larger the value, the higher the surface roughness. The energy and uniformity indexes reflect the consistency of the surface texture, with higher values ​​indicating more uniform polishing; When the calculated roughness is higher than the preset threshold, it indicates that there is still slight uneven wear or grinding residue on the surface of the phosphor bronze ball. At this time, the polishing parameters are adjusted and grinding is continued.

3. The method for constructing a phosphor bronze ball production polishing detection model according to claim 1, characterized in that: The brightness histogram analysis method is used to calculate the finish uniformity of phosphor bronze balls, including: The brightness distribution of the phosphor bronze ball area in the optical image is counted, and the standard deviation, kurtosis and skewness of the brightness histogram are calculated to evaluate whether the surface finish is uniform; The standard deviation of the brightness histogram reflects the degree of dispersion of the brightness distribution. The lower the value, the more uniform the surface finish. The kurtosis indicates the concentration of the image brightness. The higher the value, the better the polishing uniformity. The skewness is used to judge the symmetry of the brightness distribution. The lower the value, the more uniform the surface finish. If the finish is lower than the preset threshold, it means that there are still local rough areas or grinding residues on the surface of the phosphor bronze ball, and polishing needs to be continued; When the test results show that the surface roughness of the phosphor bronze ball is high or the finish is low, the polishing parameters are adjusted to continue polishing.

4. A phosphor bronze ball production polishing detection model construction system, characterized in that: The system includes the following modules: A data acquisition module is used to obtain an optical image of the phosphor bronze ball on the top of the polishing barrel using an industrial camera, and to obtain a spectral image of the top of the polishing barrel using a short-wave infrared hyperspectral camera; The calibration module is used to align the optical image and the spectral image in space so that the two can be analyzed in the same coordinate system; the first position of the phosphor bronze ball in the spectral image is determined by the target detection algorithm, and the corresponding second position of the phosphor bronze ball in the optical image is calculated based on the coordinate mapping model of the optical image to ensure the consistency of the same phosphor bronze ball in the two data channels; a first judgment module, configured to calculate the roundness of the phosphor bronze ball at the second position by using an ellipse fitting algorithm, and judge the geometric integrity of the phosphor bronze ball based on a roundness threshold, and when the roundness of the phosphor bronze ball does not meet the standard, adjust the polishing parameters and continue grinding; The second judgment module is used to calculate the surface roughness index of the phosphor bronze ball at the second position by using the gray level co-occurrence matrix analysis method to evaluate whether the polishing is uniform; the brightness histogram analysis method is used to calculate the smoothness uniformity of the phosphor bronze ball to determine whether it meets the polishing standard; when the smoothness of the phosphor bronze ball is lower than the preset threshold, the polishing parameters are adjusted and polishing is continued, otherwise the polishing is stopped and the detection result is output The positions of the industrial camera and the short-wave infrared hyperspectral camera are relatively fixed; The spatial alignment of the optical image and the spectral image is achieved by the following modules: The intrinsic parameter calibration module is used to perform camera intrinsic parameter calibration. It uses a chessboard or calibration plate to calibrate the intrinsic parameter matrices of the optical camera and spectral camera, including focal length, principal point coordinates, and lens distortion coefficient. The external parameter calibration module is used to perform camera external parameter calibration to determine the relative position relationship between the optical camera and the spectral camera, that is, the rotation matrix and translation vector, which are used to describe the coordinate transformation of the two cameras and calculate the homography matrix between the cameras; The spatial position matching module is used to convert the pixel points in the spectral image into the coordinate system of the optical image using the calculated rotation matrix and translation vector to ensure that the spatial positions of the two match; The remapping module is used to calculate the corresponding position of each pixel in the spectral image in the optical image by using the projection transformation matrix through homography transformation and perform remapping.

5. The phosphor bronze ball production polishing detection model construction system according to claim 4, characterized in that: The gray level co-occurrence matrix analysis method is used to calculate the roughness index of the phosphor bronze ball surface through the following modules: The image preprocessing module is used to perform grayscale normalization on the optical image of the phosphor bronze ball, set a fixed pixel window, calculate the grayscale co-occurrence matrix between adjacent pixels in different directions, and quantify the surface roughness by extracting contrast, energy, uniformity, and entropy features; The grinding judgment module is used to adjust the polishing parameters and continue grinding when the calculated roughness is higher than the preset threshold, indicating that there is still slight uneven wear or grinding residue on the surface of the phosphor bronze ball.

6. The phosphor bronze ball production polishing detection model construction system according to claim 4, characterized in that: The brightness histogram analysis method is used to calculate the smoothness uniformity of phosphor bronze balls using the following modules: The brightness distribution evaluation module is used to count the brightness distribution of the phosphor bronze ball area in the optical image and calculate the standard deviation, kurtosis and skewness of the brightness histogram to evaluate whether the surface finish is uniform; Brightness histogram evaluation module, used for the standard deviation of the brightness histogram to reflect the discreteness of the brightness distribution. The lower the value, the more uniform the surface finish. The kurtosis indicates the concentration of the image brightness. The higher the value, the better the polishing uniformity. The skewness is used to judge the symmetry of the brightness distribution. The lower the value, the more uniform the surface finish. The smoothness evaluation module is used to indicate that if the smoothness is lower than the preset threshold, it indicates that there are still local rough areas or grinding residues on the surface of the phosphor bronze ball, and polishing needs to be continued; The polishing judgment module is used to adjust the polishing parameters and continue polishing when the detection results show that the surface roughness of the phosphor bronze ball is high or the finish is low.

Images