Rotating disc type ceramic chopper positioning and cleaning method based on machine vision

Abstract

The invention discloses a rotating disc type ceramic chopper positioning and cleaning method based on machine vision, and relates to the technical field of microelectronic packaging, and the method comprises the steps of image acquisition, inner contour coarse positioning, outer contour fine positioning, hole center positioning and state judgment. Through organic combination of a special imaging scheme and a visual algorithm, precise positioning of hole position coordinates of the rotating disc type ceramic chopper cleaning machine is achieved, field actual production verifies that under the condition that dirt and water stain influence exists around hole positions, the hole center position locking accuracy reaches 99.97%, and the hole center position locking precision reaches 99.97%. The problem of insufficient positioning and inserting precision in the prior art is effectively solved; meanwhile, reliable support is provided for rotating disc type high-pressure directional cleaning through accurate positioning, cleaning pertinence and efficiency are improved, product damage is reduced, linkage automation of positioning and cleaning procedures is achieved, the problems of high labor intensity, secondary pollution and the like caused by manual intervention are solved, and the production benefit and the product percent of pass are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of microelectronic packaging technology, specifically a rotary ceramic chopping tool positioning and cleaning method based on machine vision. Background Technology

[0002] As microelectronic packaging technology rapidly advances towards higher density and precision, ceramic cleavers, as core components, have effectively overcome the bottlenecks of traditional packaging processes thanks to the ultra-high hardness, excellent wear resistance, and high-temperature stability of ceramic materials such as zirconium oxide and alumina. Compared to traditional metal cleavers, ceramic cleavers maintain edge sharpness and structural integrity during long-term, high-frequency bonding operations, reducing tool wear. Simultaneously, they resist temperature fluctuations in high-temperature bonding processes, ensuring consistency in the packaging process and high product yield.

[0003] The production of ceramic choppers involves multiple processes, including sintering, machining, and laser forming. The tapered end face is highly susceptible to contaminants such as ceramic powder, cutting debris, and oil. If these impurities are not thoroughly removed, they can lead to abnormal assembly clearances, increased contact wear during use, and directly affect the strength of the encapsulation connection and the lifespan of the device. Therefore, precise cleaning of the tapered end face is an indispensable quality control step in the production of ceramic choppers.

[0004] However, existing technologies have many shortcomings: First, the positioning and insertion accuracy is insufficient. Traditional manual positioning is affected by operational errors and visual fatigue, which can easily lead to misaligned insertion and end face collisions. Automated mechanical positioning has poor adaptability and cannot meet the needs of diverse products. Second, there is an imbalance between the targeting and efficiency of cleaning. Ultrasonic cleaning has limited ability to clean gaps and dead corners and is inefficient. General high-pressure spray lacks directional design, which can easily cause incomplete cleaning or product damage. Moreover, the linkage between positioning and cleaning processes is poor. Third, the degree of automation integration is low. Positioning and insertion and cleaning are mostly independent processes, which rely on manual intervention. This results in high labor intensity, low efficiency, and the risk of secondary pollution. Existing rotary cleaning equipment and machine vision technology have not been specifically integrated for ceramic chopping knives and cannot meet the integrated requirement of "precise positioning and insertion + directional and efficient cleaning".

[0005] Based on this, a machine vision-based rotary ceramic chopping knife positioning and cleaning method is now provided, which can eliminate the drawbacks of existing devices. Summary of the Invention

[0006] The purpose of this invention is to provide a machine vision-based rotary ceramic chopping knife positioning and cleaning method to solve the problems in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A machine vision-based method for positioning and cleaning rotary ceramic chopping knives includes the following steps: S1: Image Acquisition: An imaging system consisting of a high frame rate industrial camera, a telecentric lens, and a white ring light source is used to acquire images of the holes on the rotary cleaning machine that carry ceramic cutting tools. The hole images include scenes with no material, with material, and with water stains. S2: Coarse localization of inner contour: The acquired hole position image is segmented based on the dual threshold adaptive segmentation algorithm. The target region is marked by connected component analysis, and noise is removed by two morphological opening operations to obtain the coarse localization coordinates of the inner contour of the hole position. S3: Fine positioning of outer contour: The hole position image is subjected to platform histogram equalization to enhance the edge information of the outer circle contour. After grayscale transformation, morphological processing and contour smoothing, the Huber fitting algorithm is used to perform circle fitting on the smoothed outer circle contour to obtain the accurate coordinates of the outer circle center. The platform threshold value of platform histogram equalization is 20-30. S4: Hole center positioning and status judgment: The hole center is dually positioned by combining the coarse positioning coordinates of the inner contour and the fine positioning coordinates of the outer contour. The platform threshold of the platform histogram equalization is adjusted to 70-80. The size of the white area in the image is used to determine whether a ceramic chopping tool has been inserted into the hole.

[0008] Based on the above technical solutions, the present invention also provides the following optional technical solutions: In one alternative approach: In step S2, the dual-threshold adaptive segmentation algorithm divides the image pixels into three intervals [0,T1], [T1+1,T2], and [T2+1,255] by setting two thresholds T1 and T2, and calculates T1 and T2 corresponding to the maximum variance of each interval as the optimal segmentation thresholds.

[0009] In one alternative approach: the variance calculation of the dual-threshold adaptive segmentation algorithm satisfies: MAX = W0(U U0)+W1(U U1)+W2(U U2), where W0, W1, and W2 are the proportions of the three pixel intervals, U0, U1, and U2 are the average pixel values ​​of the three pixel intervals, and U is the overall pixel average of the image. When MAX reaches its maximum value, the optimal thresholds T1 and T2 are determined.

[0010] In one alternative approach: In step S3, the morphological opening operation is an erosion operation followed by a dilation operation, expressed as follows: The opening operation of the structuring element b on a grayscale image is denoted as... Corrosion operation It can suppress small particle noise and irrelevant details in binary images, effectively eliminate the fine adhesion between objects, and perform dilation operations. It can connect targets with cracks and gaps, and thicken narrow parts.

[0011] In one alternative approach: In step S3, the grayscale transformation function for platform histogram equalization is: Where L = 0~256 is the grayscale range of the image. , This is the cumulative distribution function after platformization.

[0012] In one alternative approach: In step S3, the loss function used by the Huber fitting algorithm is: , where α is the fitting residual and δ is the residual threshold.

[0013] In one alternative approach: during platform histogram equalization, the non-normalized gray-level histogram of image f(x,y) is H(k), where k is the gray level in the image, the total number of pixels in the image is N, the gray-level range is [fmin, fmax], and the gray-level probability distribution function P(k) of f(x,y) is... .

[0014] In one alternative: the cumulative distribution function of the image f(x,y) for: .

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention achieves precise positioning of hole coordinates in a rotary ceramic cleaver cleaning machine by organically combining a dedicated imaging scheme with a visual algorithm. Actual production verification shows that even with dirt and water stains around the hole, the accuracy of hole center positioning reaches 99.97%, effectively overcoming the problem of insufficient positioning and insertion accuracy in existing technologies. Simultaneously, precise positioning provides reliable support for rotary high-pressure directional cleaning, improving cleaning targeting and efficiency, reducing product damage, and achieving automated linkage between positioning and cleaning processes. This reduces labor intensity and secondary pollution caused by manual intervention, significantly improving production efficiency and product qualification rate, and meeting the requirements of large-scale production for product quality, efficiency, and stability. Attached Figure Description

[0016] Figure 1 This is a flowchart of the algorithm of the present invention.

[0017] Figure 2 A comparison image before and after performing connected component analysis and two opening operations on the image of this invention.

[0018] Figure 3 This is a diagram showing the effect of finding the outer circle contour in the platform histogram equalization of the present invention.

[0019] Figure 4 This is an image illustrating the effect of the platform histogram equalization method in determining the presence or absence of material at the orifice position, as presented in this invention.

[0020] Figure 5 This is a schematic diagram of fitting the inner and outer contour edges using the Huber method of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0022] In one embodiment, such as Figures 1-5 As shown, a machine vision-based rotary ceramic chopping knife positioning and cleaning method includes the following steps: S1: Image Acquisition: An imaging system consisting of a high frame rate industrial camera, a telecentric lens, and a white ring light source is used to acquire images of the holes on the rotary cleaning machine that carry ceramic cutting tools. The hole images include scenes with no material, with material, and with water stains. S2: Coarse localization of inner contour: The acquired hole position image is segmented based on the dual threshold adaptive segmentation algorithm. The target region is marked by connected component analysis, and noise is removed by two morphological opening operations to obtain the coarse localization coordinates of the inner contour of the hole position. S3: Fine positioning of outer contour: The hole position image is subjected to platform histogram equalization to enhance the edge information of the outer circle contour. After grayscale transformation, morphological processing and contour smoothing, the Huber fitting algorithm is used to perform circle fitting on the smoothed outer circle contour to obtain the accurate coordinates of the outer circle center. The platform threshold value of platform histogram equalization is 20-30. S4: Hole center positioning and status judgment: The hole center is dually positioned by combining the coarse positioning coordinates of the inner contour and the fine positioning coordinates of the outer contour. The platform threshold of the platform histogram equalization is adjusted to 70-80. The size of the white area in the image is used to determine whether a ceramic chopping tool has been inserted into the hole.

[0023] The above embodiments disclose a rotary ceramic chopping knife positioning and cleaning method based on machine vision. Its specific working principle and process are as follows: The ceramic chopping blade has a bottom diameter of approximately 1.585mm, while the diameter of the hole in the rotary cleaning machine is 1.66mm. This necessitates extremely high positioning accuracy. Furthermore, there are traces of glue bonding and residual water stains near the hole. Therefore, precise positioning is required before and after the chopping blade is inserted or removed, even when the hole is obscured. Based on this, this invention uses an outer contour circle concentric with the hole for positioning. The core components include a dedicated imaging scheme and a visual algorithm, as detailed below. 1. Dedicated imaging solution: To capture clear real-time images of the holes at the bottom of the equipment, a high-frame-rate industrial camera is used for real-time image acquisition. A telecentric lens is used to ensure clear imaging of the hole positions and outer contour details. A white ring light source is used to reduce other redundant information, ensuring that high-quality hole position images can be obtained in different scenarios such as no material, with material, and with water stains.

[0024] 2. Visual Algorithms: The core objectives of the algorithm are twofold: (1) Precise positioning: Under complex backgrounds containing interference factors such as water stains, accurately identify and locate the center coordinates of the target hole position carrying the ceramic chopping knife; (2) Status judgment: Determine whether the ceramic chopping knife has been successfully inserted into the target hole position. The algorithm combines the advantages of coarse positioning of the inner contour of the hole position and fine positioning of the outer contour to achieve dual positioning of the hole center. Through the organic combination of a series of image processing and analysis operators, it ensures high robustness and high positioning accuracy in dynamic production environments. The specific process is as follows: (1) Coarse positioning of the inner contour: The goal is to quickly narrow down the processing scope, eliminate most background interference, and improve the efficiency and accuracy of subsequent algorithms.

[0025] Dual-threshold adaptive segmentation: This method is an extension of the single-threshold adaptive segmentation principle. Single-threshold segmentation uses histograms to obtain the probability of occurrence of each segmentation feature value. It divides the segmentation feature values ​​into two classes using a threshold variable, calculates the intra-class variance and inter-class variance for each class, and selects T, which maximizes the inter-class variance or minimizes the intra-class variance, as the optimal threshold. However, single-threshold segmentation has limitations, failing to effectively segment targets in complex image backgrounds. Therefore, this invention refines the segmentation interval by setting two thresholds, achieving accurate target segmentation. The segmentation interval is defined as follows:

[0026] In the formula, C1, C2, and C3 represent three pixel intervals. To maximize the variance of these three intervals, T1 and T2 represent two desired thresholds. The calculation formula is as follows:

[0027] In the formula, W0, W1, W2 represent the proportions of the three regions C1, C2, and C3 in the image, and U1, U2, and U3 represent the average pixel values ​​of the three regions C1, C2, and C3. When MAX reaches its maximum, T1 and T2 are the target values, i.e., adaptive dual thresholds.

[0028] Connected component analysis: Region labeling is performed on a binary image, and connected component attribute analysis is completed based on the connected component label threshold. A connected component is an image region consisting of foreground pixels with the same pixel value and adjacent positions. Connected component analysis involves identifying and labeling each connected component in the image. The formula for calculating the direction of the connected component label is:

[0029] In the formula, The region label direction is represented by M, the pixel area of ​​the connected region is represented by R, the image function of the connected region is represented by L, and the perimeter of the region (the sum of the pixels at the region boundary) is represented by L. These conditions are used to find connected regions in the image and assign them unique identifiers to distinguish them from other connected regions.

[0030] Morphological double opening operation: The opening operation of the structuring element b on a grayscale image is denoted as f. b is defined as: ,in The symbol ⊕ represents the erosion operation, and ⊕ represents the dilation operation. The opening operation requires both erosion and dilation. Erosion typically suppresses small particle noise and irrelevant details in binary images, effectively eliminating fine adhesion between objects. Dilation can connect objects with cracks and gaps, and thicken narrow sections. By performing erosion followed by dilation, protrusions on image object boundaries can be suppressed, narrow connections can be separated, and small isolated points, broken lines, and patchy areas can be eliminated. This makes the image object edges smoother without significantly altering the size and shape of the main object area. It can remove a large number of excessively small noise points, avoiding interference with the main edge information of the region, and ultimately obtaining coarse positioning coordinates for the initial fitting of the inner contour.

[0031] (2) Precision positioning of the outer circle contour: By identifying the outer circular contour of the bearing hole, the coordinates of the hole center are corrected and refined. The positioning accuracy is improved by leveraging the fact that the outer circular contour is generally clearer and has less interference than the inner circle. The specific process is as follows: Platform histogram equalization: an image enhancement technique based on grayscale transformation. Its basic principle is to set an upper threshold (platform threshold) for the histogram. Histograms with more pixels than the platform threshold are limited to the upper threshold, while histograms below the platform threshold remain unchanged. This improves the situation where a large number of grayscale levels are merged due to obvious peaks in the image histogram. It flexibly controls the enhancement intensity of different brightness areas, effectively adjusts image contrast or normalizes data distribution, and achieves the effect of improving image quality and sharpening edges. If the non-normalized grayscale histogram of image f(x,y) is H(k), where k is the grayscale level in the image, the total number of pixels in the image is N, and the grayscale range is [fmin, fmax], then the grayscale probability distribution function P(k) is: ; Cumulative distribution function of image f(x,y) for: ; Assuming the upper threshold of the histogram is T, flatten the grayscale histogram of the image f(x,y): ; The probability distribution after platformization is as follows: ; The cumulative distribution function of the platform histogram is: ; The grayscale transformation function for platform histogram equalization is: ; Where L is the grayscale range of the image (L=0~255); This is the grayscale value after histogram equalization by the platform. Practical experience has shown that a platform threshold T value of around 20-30 provides the best display effect for the outer contour. Subsequent processing: The image after platform histogram equalization is subjected to grayscale transformation, morphological processing and contour smoothing. Finally, the Huber fitting algorithm is used to fit the smoothed contour to a circle to obtain more accurate coordinates of the outer circle center. (3) Judgment of whether there is a chopping blade: When the platform threshold T for histogram equalization is around 70-80, after inserting a ceramic chopping knife into the hole, the knife will block the light and generate its own grayscale characteristics, resulting in a reduction in the area of ​​the black area. At the same time, the area of ​​the white area will increase due to the reflection of the knife surface or the contrast with the background. When the hole is empty, due to light reflection and other reasons, the area of ​​the black area will be very large, while the area of ​​the white area will be almost non-existent. Therefore, the size of the white area can be used to determine whether the chopping knife has been inserted.

[0032] (4) Optimization of circle fitting algorithm: After locating the approximate inner and outer circular contours of the hole, curve fitting is required. The commonly used least squares method uses minimizing the mean square error as the cost function to determine the optimal function match for the data, requiring the weighted sum of squared deviations of the observed values ​​yi to be minimized. While the value of is minimized, the actual fitted curve will inevitably contain outliers xi that deviate from the normal value. These outliers can cause significant errors when the least squares method is used for linear fitting. Therefore, this invention uses the Huber loss function to improve upon this approach. This function combines the advantages of linear regression loss and absolute value loss, enhancing the robustness of the least squares method to noise outliers. The function formula is as follows:

[0033] In the formula, α is the residual (the distance between the generated line and the discrete points in linear fitting); δ is the residual threshold, used to determine whether it is an error point. After introducing the Huber loss function, the sensitivity to outliers in the boundary curve is lower than that of the least squares fitting function. No matter how large the fitting loss is, the loss function of the least squares method always grows quadratically, while the Huber algorithm only shows linear growth of the loss function when the residual value is greater than the residual threshold δ. This can reduce the influence of outliers on the fitted circle equation, while not losing information of important contour points, resulting in more ideal fitting results for circular regions.

[0034] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A machine vision-based method for positioning and cleaning rotary ceramic chopping knives, characterized in that, Includes the following steps: S1: Image Acquisition: An imaging system consisting of a high frame rate industrial camera, a telecentric lens, and a white ring light source is used to acquire images of the holes on the rotary cleaning machine that carry ceramic cutting tools. The hole images include scenes with no material, with material, and with water stains. S2: Coarse localization of inner contour: The acquired hole position image is segmented based on the dual threshold adaptive segmentation algorithm. The target region is marked by connected component analysis, and noise is removed by two morphological opening operations to obtain the coarse localization coordinates of the inner contour of the hole position. S3: Fine positioning of outer contour: The hole position image is subjected to platform histogram equalization to enhance the edge information of the outer circle contour. After grayscale transformation, morphological processing and contour smoothing, the Huber fitting algorithm is used to perform circle fitting on the smoothed outer circle contour to obtain the accurate coordinates of the outer circle center. The platform threshold value of platform histogram equalization is 20-30. S4: Hole center positioning and status judgment: The hole center is dually positioned by combining the coarse positioning coordinates of the inner contour and the fine positioning coordinates of the outer contour. The platform threshold of the platform histogram equalization is adjusted to 70-80. The size of the white area in the image is used to determine whether a ceramic chopping tool has been inserted into the hole.

2. The method for positioning and cleaning a rotary ceramic chopping knife based on machine vision according to claim 1, characterized in that, In step S2, the dual-threshold adaptive segmentation algorithm divides the image pixels into three intervals [0,T1], [T1+1,T2], and [T2+1,255] by setting two thresholds T1 and T2, and calculates the T1 and T2 corresponding to the maximum variance of each interval as the optimal segmentation threshold.

3. The rotary ceramic chopping knife positioning and cleaning method based on machine vision according to claim 2, characterized in that, The variance calculation of the dual-threshold adaptive segmentation algorithm satisfies: MAX = W0(U U0)+W1(U U1)+W2(U U2), where W0, W1, and W2 are the proportions of the three pixel intervals, U0, U1, and U2 are the average pixel values ​​of the three pixel intervals, and U is the overall pixel average of the image. When MAX reaches its maximum value, the optimal thresholds T1 and T2 are determined.

4. The method for positioning and cleaning a rotary ceramic chopping knife based on machine vision according to claim 1, characterized in that, In step S3, the morphological opening operation is an erosion operation followed by a dilation operation, expressed as follows: The opening operation of the structuring element b on a grayscale image is denoted as... Corrosion operation It can suppress small particle noise and irrelevant details in binary images, effectively eliminate the fine adhesion between objects, and perform dilation operations. It can connect targets with cracks and gaps, and thicken narrow parts.

5. The method for positioning and cleaning a rotary ceramic chopping knife based on machine vision according to claim 1, characterized in that, In step S3, the grayscale transformation function for platform histogram equalization is: Where L = 0~256 is the grayscale range of the image. , This is the cumulative distribution function after platformization.

6. The method for positioning and cleaning a rotary ceramic chopping knife based on machine vision according to claim 1, characterized in that, In step S3, the loss function used by the Huber fitting algorithm is: , where α is the fitting residual and δ is the residual threshold.

7. The method for positioning and cleaning a rotary ceramic chopping knife based on machine vision according to claim 5, characterized in that, During the platform histogram equalization process, the non-normalized gray-level histogram of image f(x,y) is H(k), where k is the gray level in the image, the total number of pixels in the image is N, the gray-level range is [fmin, fmax], and the gray-level probability distribution function P(k) of f(x,y) is... .

8. The method for positioning and cleaning a rotary ceramic chopping knife based on machine vision according to claim 7, characterized in that, Cumulative distribution function of image f(x,y) for: .

Images