A visual inspection method for burr defects in wafer boxes

Abstract

The present invention relates to the field of image data processing technology, and in particular to a wafer box burr defect visual detection method, comprising: obtaining edge information sequences corresponding to groove areas in a wafer box image, analyzing the similarity relationship and distribution characteristics between different edge information sequences, respectively calculating the groove symmetry value of the wafer box and the groove period value of the groove area, obtaining the groove characteristic parameters of the wafer box, obtaining the burr coefficient of the wafer box according to the grayscale distribution of pixel points in the wafer box image, performing iterative morphological processing on all groove areas of the wafer box image, combining the groove characteristic parameters and burr coefficient of the wafer box after each morphological processing, drawing a groove burr curve, and performing burr defect detection according to the change of the groove detection curve. The present invention improves the detection accuracy of burr defects in wafer boxes.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular to a method for visually detecting burr defects in a wafer box. Background Art

[0002] Wafer boxes are sealed containers used to store and transport wafers in semiconductor production. They are usually made of heat-resistant and wear-resistant translucent plastics and have properties such as sealing and anti-static properties. They are designed to minimize the risk of wafers being contaminated and scratched during transportation and to ensure the integrity of the wafers as much as possible. However, during the manufacturing process of wafer boxes, burrs may be generated on the edges of the grooves inside the wafer boxes due to process defects or abnormal operation of manufacturing equipment. Burrs can cause wafers to be scratched during storage, posing a potential threat to wafer quality.

[0003] Therefore, in order to ensure the storage safety of wafers in the wafer box, non-destructive testing of the wafer box is usually performed in combination with machine vision technology. In the prior art, non-destructive testing of wafer boxes can use machine vision algorithms to detect burr defects in the wafer box through images. However, since the burrs in the grooves of the wafer box are small, they are not easy to detect in the images obtained by machine vision. The fixed morphological processing algorithm in the prior art is prone to missed detection of small burrs, resulting in low accuracy in the detection results of burr defects in the wafer box, which cannot effectively meet the requirements for wafer box quality in industrial production. Summary of the invention

[0004] The invention provides a wafer box burr defect visual detection method to solve the existing problems.

[0005] A wafer box burr defect visual detection method of the present invention adopts the following technical solution:

[0006] An embodiment of the present invention provides a method for visually detecting burr defects in a wafer box, the method comprising the following steps:

[0007] Acquire a wafer box image, wherein both sides of the wafer box image contain groove areas;

[0008] The edge information sequences corresponding to the groove areas on each side are obtained respectively, the similarity relationship between different edge information sequences and the distribution characteristics of the elements in the sequences are analyzed, the groove symmetry value of the wafer box and the groove period value of the groove area are calculated respectively, and the groove characteristic parameters of the wafer box are obtained by combining the groove symmetry value and the groove period value of the wafer box;

[0009] According to the grayscale distribution of the pixels in the wafer box image, the burr coefficient of the wafer box is obtained;

[0010] Iterative morphological processing is performed on all groove areas of the wafer box image. In each morphological processing process, burr feature parameters are obtained according to the groove feature parameters and burr coefficients of the wafer box; and a groove burr curve is drawn through the burr feature parameters corresponding to all sub-morphological processing processes;

[0011] According to the change of the groove detection curve, the wafer box burr defect detection result is obtained.

[0012] Furthermore, the specific method of respectively obtaining the edge information sequence corresponding to the groove area on each side includes:

[0013] Use Canny edge detection to perform edge detection on each groove area to obtain the edge image corresponding to each groove area;

[0014] The edge image of each groove area is chain-coded using the 8-connected chain code to obtain the chain code sequence corresponding to each groove area, which is recorded as the edge information sequence of the groove area.

[0015] Furthermore, the similarity relationship between different edge information sequences and the distribution characteristics of elements in the sequences are analyzed to calculate the groove symmetry value of the wafer box and the groove period value of the groove area, respectively, including the specific method of:

[0016] According to the similarity relationship and distribution difference between the edge information sequences of the groove areas on both sides, the groove symmetry value of the wafer box is obtained;

[0017] For the groove area on any side of the wafer box, the edge information sequence of the groove area is decomposed to obtain the residual term of the edge information sequence; and the groove period value of the groove area is calculated according to the distribution characteristics of the data points in the residual term of the edge information sequence.

[0018] Furthermore, the method for obtaining the distribution difference between the edge information sequences of the groove areas on both sides includes:

[0019] The edge direction distribution volatility in each edge information sequence is obtained, and the difference in the edge direction distribution volatility is used as the distribution difference between the edge information sequences.

[0020] Furthermore, the distribution characteristics of the data points in the residual term of the edge information sequence include: the variance of the data points in the residual term of the edge information sequence and the number of data points whose values ​​are greater than an average residual parameter, and the average residual parameter is the mean of the values ​​of all data points in the residual term of the edge information sequence.

[0021] Furthermore, the burr coefficient of the wafer box is obtained according to the grayscale distribution of the pixels in the wafer box image, including the specific method of:

[0022] The second-order moment of the grayscale value corresponding to any edge pixel point in the groove area in the wafer box image is obtained, recorded as the grayscale distribution parameter of the edge pixel point, and the grayscale distribution parameters of all edge pixel points in the groove areas on both sides are integrated to obtain the burr coefficient of the wafer box.

[0023] Furthermore, the iterative morphological processing of all the groove regions of the wafer box image includes the following specific methods:

[0024] A morphological convolution kernel is constructed within a preset iteration range, and convolution kernels of different sizes are used to perform morphological corrosion processing on all groove areas respectively, so as to obtain the groove area after corrosion processing, which is recorded as the groove corrosion image under the corresponding convolution kernel size.

[0025] Furthermore, in each morphological processing process, the burr feature parameters are obtained according to the groove feature parameters and the burr coefficient of the wafer box, including the specific method of:

[0026] According to the wafer box corrosion image under each convolution kernel size, the groove characteristic parameters and burr coefficient of the wafer box under the corresponding convolution kernel size are obtained;

[0027] According to the groove feature parameters and burr coefficient of the wafer box under each convolution kernel size, the burr feature parameters under the corresponding convolution kernel size are obtained, the groove feature parameters are directly proportional to the burr feature parameters, and the burr coefficient is inversely proportional to the burr feature parameters.

[0028] Furthermore, the method of drawing the groove burr curve by using the burr feature parameters corresponding to all the sub-morphological processing processes includes the following specific methods:

[0029] A two-dimensional rectangular coordinate system is constructed, the convolution kernel size is used as the horizontal axis of the two-dimensional rectangular coordinate system, the burr feature parameters under different convolution kernel sizes are used as the vertical axis of the two-dimensional rectangular coordinate system, and a scatter plot of the burr feature parameters under all convolution kernel sizes in the two-dimensional rectangular coordinate system is obtained. The least squares method is used to perform curve fitting on all scattered points to obtain the groove burr curve.

[0030] Furthermore, the wafer box burr defect detection result is obtained according to the change of the groove detection curve, and the specific method includes:

[0031] The second-order derivatives of all data points in the groove burr curve are obtained. When the value of the second-order derivative corresponding to a data point is greater than 0, a burr defect exists in the wafer box, and the wafer box with the burr defect is marked.

[0032] The beneficial effect of the technical solution of the present invention is: the embodiment of the present invention takes into account that the grooves on both sides of the wafer box are usually symmetrical to each other, so the distribution of the grooves in the wafer box is symmetrical on both sides and the grooves on either side show a periodic distribution feature, and when burr defects appear between the grooves of the wafer box, it will cause the above-mentioned characteristics corresponding to the grooves of the wafer box to be destroyed, so the embodiment of the present invention uses an edge information sequence to quantify the symmetry and periodicity of the groove area. The embodiment of the present invention takes into account that when there are burrs in the groove area, after multiple morphological processings have been performed to remove the area corresponding to the burr defect, the degree to which the symmetry features, periodic features and grayscale features corresponding to the grooves on both sides of the wafer box are affected by the burrs will be minimized. Therefore, through iterative morphological processing, the similarity relationship and distribution characteristics between different edge information sequences are analyzed during the iterative processing, and the groove symmetry value and groove period value of the wafer box are calculated. The groove characteristic parameters of the wafer box can be accurately described, the burr coefficient is obtained in combination with the grayscale distribution, and the groove burr curve is quickly obtained based on the curve fitting method. According to the changes in the groove detection curve, it is possible to accurately judge whether the wafer box has a burr defect, thereby improving the detection accuracy of burr defects in the wafer box. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 A flowchart of the steps of a wafer box burr defect visual detection method of the present invention;

[0035] Figure 2 This is a schematic diagram of the groove burr curve. DETAILED DESCRIPTION

[0036] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a wafer box burr defect visual detection method proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0037] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0038] The specific scheme of the wafer box burr defect visual detection method provided by the present invention is described in detail below with reference to the accompanying drawings.

[0039] See also Figure 1 , which shows a flowchart of a method for visually detecting burr defects of a wafer box provided by an embodiment of the present invention, the method comprising the following steps:

[0040] Step S001: Acquire a wafer box image, wherein both sides of the wafer box image contain groove areas.

[0041] It should be noted that the wafer box is usually designed with multiple grooves to fix and protect the wafer. However, due to the design limitations of the wafer box mold, material fluidity differences and other factors, the wafer box is usually made by injection molding. If the mold has defects in the manufacturing process or is not well maintained during use, the parting surface, slider or other parts of the mold may be worn or accumulate impurities, resulting in burrs during the wafer box molding process; during the injection molding process, if the injection pressure, temperature, cooling speed and other parameters are not set properly, or if there are impurities or poor uniformity in the plastic material, burrs may appear on the edges of the grooves where the wafers are placed inside the wafer box. Although these burrs are small, they will cause defects such as scratches on the wafer, affecting the performance and yield of the wafer. Therefore, non-contact burr detection of the grooves on both sides of the wafer box through machine vision technology is a necessary means to ensure the quality of the wafer box.

[0042] Specifically, in order to implement a wafer box burr defect visual detection method proposed in this embodiment, it is first necessary to collect a wafer box image. The specific process is as follows:

[0043] First, place the wafer box on the conveyor and use a CCD camera to take a picture from above to obtain the wafer box image, such as Figure 2 Schematic diagram of wafer box.

[0044] Then, the obtained wafer box image is subjected to grayscale processing and Gaussian filtering.

[0045] Finally, the semantic segmentation algorithm is used to obtain the areas corresponding to the grooves on both sides of the wafer box image, which are recorded as groove areas. The groove areas on both sides of the wafer box are called the first groove area and the second groove area, respectively.

[0046] So far, the wafer box image and the groove area in the wafer box image are obtained through the above method.

[0047] Step S002: respectively obtain the edge information sequences corresponding to the groove areas on each side, analyze the similarity relationship between different edge information sequences and the distribution characteristics of the elements in the sequences, respectively calculate the groove symmetry value of the wafer box and the groove period value of the groove area, and combine the groove symmetry value and groove period value of the wafer box to obtain the groove characteristic parameters of the wafer box.

[0048] It should be noted that the size of the grooves inside the wafer box and the spacing between the grooves are set according to production standards. At the same time, in order to facilitate the placement of wafers, the grooves on both sides of the wafer box are usually symmetrical to each other. Therefore, the distribution of the grooves in the wafer box is symmetrical on both sides and the grooves on either side show a periodic distribution feature. When burr defects appear between the grooves of the wafer box, the above-mentioned features corresponding to the grooves of the wafer box will be destroyed. Therefore, this embodiment chooses to analyze the distribution characteristics of the grooves of the wafer box in the image and extract the corresponding edge information to quantify the symmetry characteristics and periodic characteristics of the grooves, so as to further quantify the overall characteristics of the grooves of the wafer box and obtain the groove characteristic parameters of the wafer box.

[0049] Specifically, in step 2.1, edge information sequences corresponding to the groove regions on each side are obtained respectively.

[0050] Preferably, in an embodiment of the present invention, edge information sequences corresponding to the groove areas in the wafer box image are obtained, and the specific acquisition method is: use Canny edge detection to perform edge detection on each groove area respectively, and obtain the edge image corresponding to each groove area; use 8-connected chain code to perform chain code encoding on the edge image of each groove area, and obtain the chain code sequence corresponding to each groove area, which is recorded as the edge information sequence of the groove area.

[0051] It should be noted that the corresponding edge information in the groove area of ​​the wafer box can effectively reflect the groove profile characteristics. When burrs appear between the grooves, the burrs will also appear in the groove profile. By traversing the edges of the groove area through the 8-connected chain code, the profile characteristics of the groove area can be accurately quantified.

[0052] Step 2.2, analyze the similarity relationship between different edge information sequences and the distribution characteristics of the elements in the sequence, and calculate the groove symmetry value of the wafer box and the groove period value of the groove area respectively.

[0053] Preferably, in one embodiment of the present invention, since the groove areas on both sides of a normal wafer box are usually symmetrical to each other and the grooves on each side are periodically distributed, their corresponding profile features are also similar. This embodiment uses the similarity relationship and distribution characteristics between the edge information sequences that can describe the groove profile features in the groove area to quantify the symmetry features between the groove areas and the periodic features of each groove area. According to the similarity relationship and distribution difference between the edge information sequences of the groove areas on both sides, the groove symmetry value of the wafer box is obtained.

[0054] It should be noted that, under normal circumstances, that is, when there are no burr defects in the grooves of the wafer box, the grooves of the wafer box present a periodic regular distribution, and the corresponding edge information sequence is also a periodically changing one-dimensional sequence. When a burr defect appears in the groove, the chain code value in the edge information sequence obtained by chain code encoding will change, so that the periodic distribution characteristics of the groove are affected. Therefore, in order to facilitate the quantification of the influence of the burr on the periodic distribution characteristics of the groove, this embodiment performs STL time series decomposition on the edge information sequence. Since the grooves in the wafer box are of uniform size and have a strong periodic regularity, when a burr defect occurs, the influence of the burr on the edge information sequence is usually manifested in the corresponding residual term. Therefore, in this embodiment, the distribution characteristics of the data points in the residual term are obtained to quantitatively describe the degree of burrs in the groove area.

[0055] For the groove area on any side of the wafer box, the edge information sequence of the groove area is decomposed to obtain the residual term of the edge information sequence; and the groove period value of the groove area is calculated according to the distribution characteristics of the data points in the residual term of the edge information sequence.

[0056] Preferably, the method for obtaining the distribution difference between edge information sequences of groove regions on both sides includes: obtaining edge direction distribution volatility in each edge information sequence, and taking the difference in edge direction distribution volatility as the distribution difference between edge information sequences.

[0057] Preferably, the distribution characteristics of the data points in the residual term of the edge information sequence include: the variance of the data points in the residual term of the edge information sequence and the number of data points whose values ​​are greater than an average residual parameter, and the average residual parameter is the mean of the values ​​of all data points in the residual term of the edge information sequence.

[0058] As an embodiment, since the chain code information can characterize the direction change on the edge, the specific calculation method of the groove symmetry value of the wafer box in the embodiment of the present invention is:

[0059]

[0060] In the formula, Indicates the groove symmetry value of the wafer box; An edge information sequence representing a first groove region of a wafer box; A sequence of edge information representing a second groove region of the wafer box; Indicates the DWT distance between edge information sequences; represents the variance of the chain code value in the edge information sequence of the first groove area; represents the variance of the chain code value in the edge information sequence of the second groove area; Represents the absolute value symbol.

[0061] It should be noted that It indicates the DTW distance between the edge information sequences of the groove areas on both sides of the wafer box obtained by the DTW (Dynamic Time Warping) algorithm. The distance reflects the similarity relationship between the edge information sequences. The smaller the distance, the higher the similarity between the edge information sequences, which further indicates that the degree of symmetry between the edges of the corresponding grooves is higher. In addition, It reflects the difference in the discreteness of the chain code values ​​in the edge information sequences corresponding to the groove areas on both sides of the wafer box, that is, the difference in the volatility of the edge direction distribution, and characterizes the difference in the distribution direction between the edge information sequences of the groove areas on both sides, that is, it represents the similarity relationship about the volatility of the edge direction distribution. The smaller the difference, the more similar the discreteness of the edge information sequences corresponding to the groove areas on both sides is, and the higher the symmetry relationship of the grooves on both sides of the wafer box is. Therefore, by obtaining the DTW distance between the edge information sequences and the difference in the discreteness of the chain code values ​​in the edge information sequences, the similarity relationship between the corresponding edges of the corresponding groove areas is comprehensively reflected. The higher the similarity relationship, the higher the groove symmetry value, indicating that the symmetry feature between the grooves on both sides of the wafer box is better, indicating that there are no burrs in the grooves or the impact of burr defects is smaller. The DTW algorithm is a well-known technology and will not be repeated in this embodiment.

[0062] As an embodiment, the specific calculation method of the groove period value of the groove region is:

[0063]

[0064] In the formula, represents the groove period value of the groove area, Represents the variance of the data points in the residual term of the edge information sequence corresponding to the groove area; Indicates the number of data points in the residual term of the edge information sequence corresponding to the groove area whose values ​​are greater than the average residual parameter.

[0065] The average residual parameter is the mean value of all data point values ​​in the residual term of the edge information sequence.

[0066] In one embodiment of the present invention, the STL time series decomposition algorithm is used to decompose the edge information sequence of the groove area to obtain the residual term.

[0067] Step 2.3, combining the groove symmetry value and the groove period value of the wafer box to obtain the groove characteristic parameters of the wafer box.

[0068] As an embodiment, the specific calculation method of the groove characteristic parameters of the wafer box is:

[0069]

[0070] Where Z represents the trench characteristic parameter of the wafer box, Indicates the trench symmetry value of the wafer box, represents the groove period value of the first groove area of ​​the wafer box, Indicates the groove period value of the second groove area of ​​the wafer box.

[0071] It should be noted that since the grooves of the wafer box have a certain symmetry and periodicity in structure during production design, in order to facilitate the subsequent detection of burr defects between the grooves of the wafer box according to the influence of the burrs on the symmetry and periodicity of the grooves of the wafer box in this embodiment, this embodiment combines the groove symmetry value of the wafer box and the corresponding groove period value of the groove areas on both sides to comprehensively describe the structural features corresponding to the grooves of the wafer box.

[0072] So far, the groove characteristic parameters of the wafer box are obtained through the above method.

[0073] Step S003: Obtaining the burr coefficient of the wafer box according to the grayscale distribution of the pixels in the wafer box image.

[0074] It should be noted that since the burrs are formed by the protrusions or accumulation of plastic products, the corresponding area can reflect more light and has a larger contrast than the normal area. Based on this feature, the grayscale distribution between pixels in the edge image after corrosion treatment is analyzed to obtain the burr coefficient of the wafer box.

[0075] Preferably, in one embodiment of the present invention, the second-order moment of the grayscale value corresponding to any edge pixel point in the groove area in the wafer box image is obtained and recorded as the grayscale distribution parameter of the edge pixel point, and the grayscale distribution parameters of all edge pixel points in the groove areas on both sides are fused to obtain the burr coefficient of the wafer box.

[0076] As an embodiment, the specific calculation method of the burr coefficient of the wafer box is:

[0077]

[0078] in, represents the burr coefficient of the wafer box; N is the number of edge pixels in the groove area on both sides, Indicates the first Grayscale distribution parameters of edge pixels.

[0079] It should be noted that, since under normal circumstances the transmittance of the burr is different from that of the groove part, that is, the grayscale value of the pixel corresponding to the burr defect in the wafer box image is different from the grayscale value of the pixel corresponding to the groove part, this embodiment obtains the grayscale distribution parameters of the edge pixel points to reflect the grayscale distribution of the edge pixel points. Therefore, the pixel points corresponding to the burr defect usually have a higher grayscale contrast, that is, there is a large difference in grayscale value between the burr pixel points and the groove edge pixel points. The second-order grayscale moment can be used to describe this complex grayscale distribution. The larger its value, the greater the difference in grayscale value, and the greater the possibility that the area contains burrs. Therefore, the larger the burr coefficient of the wafer box, the greater the probability that there are burrs in the groove area of ​​the wafer box.

[0080] So far, the burr coefficient of the wafer box is obtained through the above method.

[0081] Step S004: performing iterative morphological processing on all groove areas of the wafer box image. In each morphological processing process, obtaining burr feature parameters according to the groove feature parameters and burr coefficient of the wafer box; and drawing the groove burr curve through the burr feature parameters corresponding to all morphological processing processes.

[0082] It should be noted that since the burr defects appearing in the groove area of ​​the wafer box are usually relatively small, they are very susceptible to the influence of morphological processing, causing changes in the performance in the image before and after the processing. Therefore, in order to reflect the possibility of burr defects in the groove area by the changes in the structural characteristics of the groove area of ​​the wafer box before and after the morphological processing, this embodiment chooses to use the corrosion operation in morphology to perform multiple morphological processing on the groove area of ​​the wafer box. When there are burrs in the groove area, after multiple morphological processings, the corresponding area of ​​the burr defect is removed, and the degree to which the symmetry characteristics, periodic characteristics and grayscale characteristics corresponding to the grooves on both sides of the wafer box are affected by the burrs will be minimized. Therefore, based on this change feature, this embodiment draws a groove burr curve according to this change to reflect the corresponding changes in the influence of the burr defects on the groove areas on both sides of the wafer box.

[0083] In step 4.1, iterative morphological processing is performed on all the groove regions of the wafer box image.

[0084] As a preferred embodiment, the specific method is:

[0085] A morphological convolution kernel is constructed within a preset iteration range, and convolution kernels of different sizes are used to perform morphological corrosion processing on all groove areas respectively, so as to obtain the groove area after corrosion processing, which is recorded as the groove corrosion image under the corresponding convolution kernel size.

[0086] It should be noted that the iteration range is preset based on experience. , the iteration step length is 2, and then morphological convolution kernels of different sizes are obtained. The specific adjustment can be made according to the actual situation, and this embodiment does not make any specific limitation.

[0087] Optionally, in another embodiment of the present invention, morphological corrosion processing is performed on all groove areas according to a preset fixed-size morphological convolution kernel, and morphological corrosion processing is continued based on the result of one morphological corrosion processing to obtain a groove corrosion image after several morphological corrosion processing.

[0088] Step 4.2, in each morphological processing process, the burr feature parameters are obtained according to the groove feature parameters and the burr coefficient of the wafer box.

[0089] In the iterative morphological erosion process of changing the size of the morphological convolution kernel, as a preferred embodiment, the specific method is:

[0090] First, using the method in step S002 and step S003, according to the wafer box corrosion image under each convolution kernel size, the groove feature parameters and burr coefficient of the wafer box under the corresponding convolution kernel size are obtained.

[0091] Then, according to the groove feature parameters and burr coefficient of the wafer box under each convolution kernel size, the burr feature parameters under the corresponding convolution kernel size are obtained, the groove feature parameters are directly proportional to the burr feature parameters, and the burr coefficient is inversely proportional to the burr feature parameters.

[0092] As an embodiment, the specific calculation method of the burr characteristic parameter is:

[0093]

[0094] in, Indicates that the convolution kernel size is The burr characteristic parameters when Indicates that the convolution kernel size is When , the trench characteristic parameters of the wafer box; Indicates that the convolution kernel size is The burr coefficient of the wafer box when .

[0095] It should be noted that the burr feature parameter is used to describe the degree to which the groove area of ​​the wafer box image is not affected by burr defects after convolution kernel corrosion treatment. When the groove feature parameter is larger, it means that the periodicity and symmetry of the groove areas on both sides of the wafer box are less destroyed by the burrs, and the probability of containing burr defects in the groove area is smaller, and the burr feature parameter is larger; when the burr coefficient is smaller, it means that the probability of containing burrs in the groove areas on both sides of the round box is smaller, and the burr feature parameter is larger.

[0096] Optionally, in the process of iterative morphological corrosion using a preset fixed-size morphological convolution kernel, as a preferred embodiment, the method for obtaining the burr feature parameters is: according to the corrosion image of the wafer box after each morphological corrosion, the groove feature parameters and the burr coefficient of the wafer box under the corresponding number of morphological corrosion are obtained; according to the groove feature parameters and the burr coefficient of the corresponding wafer box after each morphological corrosion, the burr feature parameters after each morphological corrosion are obtained, the groove feature parameters are directly proportional to the burr feature parameters, and the burr coefficient is inversely proportional to the burr feature parameters.

[0097] Step 4.3, drawing a groove burr curve by using the burr feature parameters corresponding to all sub-morphological processing processes, as an embodiment, the specific method is:

[0098] A two-dimensional rectangular coordinate system is constructed, the convolution kernel size is used as the horizontal axis of the two-dimensional rectangular coordinate system, and the burr feature parameters under different convolution kernel sizes are used as the vertical axis of the two-dimensional rectangular coordinate system. A scatter plot of the burr feature parameters under all convolution kernel sizes in the two-dimensional rectangular coordinate system is obtained, and a curve fitting is performed on all scattered points using the least squares method to obtain a groove burr curve, such as Figure 2 Shown is a schematic diagram of the groove burr curve.

[0099] Optionally, in another embodiment of the present invention, in the process of drawing the groove burr curve using the burr characteristic parameters after all morphological etchings, the horizontal axis of the two-dimensional rectangular coordinate system is the number of morphological etchings.

[0100] So far, the groove detection curve of the wafer box is obtained through the above method.

[0101] Step S005: Obtaining wafer box burr defect detection results according to the change of the groove detection curve.

[0102] It should be noted that, in the process of etching the wafer box image using the convolution kernel with iteratively increased size, the burr defects in the wafer box image will be gradually corroded and removed, and a burr feature parameter corresponds to each etching operation. When there are no burrs in the groove areas on both sides of the wafer box, the burr feature parameters corresponding to convolution kernels of different sizes are relatively consistent in size. When there are burrs in the groove areas on both sides of the wafer box, as the size of the convolution kernel changes, at the beginning, the pixel points corresponding to the burr defect are not completely corroded, and the corresponding structural features of the groove areas on both sides of the wafer box are still affected by the burr defects, and the corresponding burr feature parameters are in a stable state and have a small value. As the convolution kernel becomes larger, the pixel points corresponding to the burr defect are completely corroded, and the corresponding structural features of the groove areas on both sides of the wafer box are no longer affected by the burr defects, then the value of the burr feature parameter will suddenly increase. Therefore, this embodiment can determine whether there are burr defects in the groove area of ​​the wafer box according to the change of the groove detection curve to ensure the production quality of the wafer box.

[0103] Preferably, in an embodiment of the present invention, the second-order derivatives of all data points in the groove burr curve are obtained. When there is a data point whose corresponding second-order derivative value is greater than 0, there is a burr defect in the wafer box, and the wafer box with the burr defect is marked.

[0104] At this point, this embodiment is completed.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for visually detecting burr defects in wafer boxes, characterized in that: The method comprises the following steps: Acquire a wafer box image, wherein both sides of the wafer box image contain groove areas; The edge information sequences corresponding to the groove areas on each side are obtained respectively, the similarity relationship between different edge information sequences and the distribution characteristics of the elements in the sequences are analyzed, the groove symmetry value of the wafer box and the groove period value of the groove area are calculated respectively, and the groove characteristic parameters of the wafer box are obtained by combining the groove symmetry value and the groove period value of the wafer box; According to the grayscale distribution of the pixels in the wafer box image, the burr coefficient of the wafer box is obtained; Iterative morphological processing is performed on all groove areas of the wafer box image. In each morphological processing process, burr feature parameters are obtained according to the groove feature parameters and burr coefficients of the wafer box; and a groove burr curve is drawn through the burr feature parameters corresponding to all sub-morphological processing processes; According to the change of the groove detection curve, the wafer box burr defect detection result is obtained.

2. According to claim 1, a wafer box burr defect visual detection method is characterized in that: The specific method of respectively obtaining the edge information sequence corresponding to the groove area on each side includes: Use Canny edge detection to perform edge detection on each groove area to obtain the edge image corresponding to each groove area; The edge image of each groove area is chain-coded using the 8-connected chain code to obtain the chain code sequence corresponding to each groove area, which is recorded as the edge information sequence of the groove area.

3. According to claim 1, a wafer box burr defect visual detection method is characterized in that: The specific method of analyzing the similarity relationship between different edge information sequences and the distribution characteristics of the elements in the sequences and respectively calculating the groove symmetry value of the wafer box and the groove period value of the groove area includes: According to the similarity relationship and distribution difference between the edge information sequences of the groove areas on both sides, the groove symmetry value of the wafer box is obtained; For the groove area on any side of the wafer box, decomposing the edge information sequence of the groove area to obtain the residual term of the edge information sequence; According to the distribution characteristics of the data points in the residual term of the edge information sequence, the groove period value of the groove area is calculated.

4. A wafer box burr defect visual detection method according to claim 3, characterized in that: The method for obtaining the distribution difference between the edge information sequences of the groove areas on both sides includes: The edge direction distribution volatility in each edge information sequence is obtained, and the difference in the edge direction distribution volatility is used as the distribution difference between the edge information sequences.

5. A wafer box burr defect visual detection method according to claim 3, characterized in that: The distribution characteristics of the data points in the residual term of the edge information sequence include: the variance of the data points in the residual term of the edge information sequence and the number of data points whose values ​​are greater than an average residual parameter, wherein the average residual parameter is the mean of the values ​​of all data points in the residual term of the edge information sequence.

6. A wafer box burr defect visual inspection method according to claim 1, characterized in that: The specific method of obtaining the burr coefficient of the wafer box according to the grayscale distribution of the pixel points in the wafer box image is as follows: The second-order moment of the grayscale value corresponding to any edge pixel point in the groove area in the wafer box image is obtained, recorded as the grayscale distribution parameter of the edge pixel point, and the grayscale distribution parameters of all edge pixel points in the groove areas on both sides are integrated to obtain the burr coefficient of the wafer box.

7. A wafer box burr defect visual detection method according to claim 1, characterized in that: The iterative morphological processing of all the groove regions of the wafer box image includes the following specific methods: A morphological convolution kernel is constructed within a preset iteration range, and convolution kernels of different sizes are used to perform morphological corrosion processing on all groove areas respectively, so as to obtain the groove area after corrosion processing, which is recorded as the groove corrosion image under the corresponding convolution kernel size.

8. A wafer box burr defect visual detection method according to claim 7, characterized in that: In each morphological processing process, the burr characteristic parameters are obtained according to the groove characteristic parameters and the burr coefficient of the wafer box, and the specific method includes: According to the wafer box corrosion image under each convolution kernel size, the groove characteristic parameters and burr coefficient of the wafer box under the corresponding convolution kernel size are obtained; According to the groove feature parameters and burr coefficient of the wafer box under each convolution kernel size, the burr feature parameters under the corresponding convolution kernel size are obtained, the groove feature parameters are directly proportional to the burr feature parameters, and the burr coefficient is inversely proportional to the burr feature parameters.

9. A wafer box burr defect visual inspection method according to claim 7, characterized in that: The method of drawing the groove burr curve by using the burr feature parameters corresponding to all the sub-morphological processing processes includes: A two-dimensional rectangular coordinate system is constructed, the convolution kernel size is used as the horizontal axis of the two-dimensional rectangular coordinate system, the burr feature parameters under different convolution kernel sizes are used as the vertical axis of the two-dimensional rectangular coordinate system, and a scatter plot of the burr feature parameters under all convolution kernel sizes in the two-dimensional rectangular coordinate system is obtained. The least squares method is used to perform curve fitting on all scattered points to obtain the groove burr curve.

10. The method for visually inspecting burr defects in wafer boxes according to claim 1, characterized in that: The specific method of obtaining the wafer box burr defect detection result according to the change of the groove detection curve is as follows: The second-order derivatives of all data points in the groove burr curve are obtained. When the value of the second-order derivative corresponding to a data point is greater than 0, a burr defect exists in the wafer box, and the wafer box with the burr defect is marked.

Images