Wafer polishing status detection method based on machine vision

Abstract

The present invention relates to the field of image processing technology, and more specifically, to a method for detecting the polishing state of a wafer based on machine vision, comprising: obtaining an interference image of a wafer; dividing the interference image into equal parts to obtain a plurality of sub-images, and calculating the roughness of a wafer region corresponding to each sub-image and the thickness of a wafer region corresponding to each sub-image; obtaining a grayscale image of the interference image, and performing binary segmentation on the grayscale image to obtain a weak edge bright stripe region, and obtaining a strong edge bright stripe region based on an adaptive region growth method. The present invention divides the interference image of the wafer into a plurality of sub-images of equal size, implements adaptive growth region segmentation of the sub-image based on the roughness of each sub-image, extracts strong edge features, and calculates the wafer polishing accuracy based on the thickness distribution difference of the extracted strong and weak edge features of the interference bright stripes, thereby realizing accurate detection of the polishing state.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and more specifically, to a method for detecting a wafer polishing state based on machine vision. Background Art

[0002] Wafer is one of the most important raw materials of LED, the light-emitting component of LED, the core part of LED. The quality of wafer will directly determine the performance of LED. Wafer is composed of III and V group compound semiconductor materials. During LED packaging, the incoming wafers are neatly arranged on the wafer film. The polishing of semiconductor material wafers is an important process of semiconductor wafer processing that uses chemical polishing, mechanical polishing or chemical mechanical polishing to remove the mechanical damage layer on the surface of the wafer and present a mirror surface. In the preparation process of semiconductor wafers, mechanical methods are often used to cut semiconductors, and the surface of the cut wafers is coarsely ground, finely ground, and chemically and mechanically polished to meet the requirements of epitaxy. Two of the tests require the testing agency to perform thickness detection and roughness detection on the surface of the semiconductor wafer.

[0003] In the related technology, for example, the Chinese patent document with the announcement number CN220288539U discloses a semiconductor wafer thickness detection mechanism. The patent document uses a scale and a pointer to facilitate personnel to observe the vertical descending distance of the fixed plate driven by the cylinder, and control the cylinder to adjust the descending distance to prevent the pad from descending too much and squeezing the semiconductor wafer too hard and causing damage. It can also control the clamping force of the clamping plate on the semiconductor wafer.

[0004] At present, the thickness and roughness measurement of wafers in different polishing states are divided into contact measurement and non-contact measurement. In contact measurement, polishing needs to be stopped for detection. The sliding of the stylus tip will leave defects on the wafer to be measured, and it is impossible to measure grooves smaller than the radius of the stylus tip, resulting in incorrect identification of the wafer polishing state and inaccurate detection of the wafer polishing state. Summary of the invention

[0005] The present invention provides a method for detecting the polishing state of a wafer based on machine vision, aiming to solve the problem in the related art that during contact measurement, polishing needs to be stopped for detection, the sliding of the stylus tip will leave defects on the wafer to be measured, and grooves smaller than the radius of the stylus tip cannot be measured, resulting in incorrect identification of the polishing state of the wafer and inaccurate detection of the polishing state of the wafer.

[0006] The present invention provides a wafer polishing state detection method based on machine vision, comprising: obtaining an interference image of a wafer; dividing the interference image into equal parts to obtain a plurality of sub-images, and calculating the roughness of a wafer region corresponding to each sub-image and the thickness of a wafer region corresponding to each sub-image; obtaining a grayscale image of the sub-image, and performing binary segmentation on the grayscale image to obtain a weak edge bright stripe region, and obtaining a strong edge bright stripe region based on an adaptive region growing method, and calculating the polishing accuracy of the sub-image , the calculation formula is: ; In the formula, is the thickness of the chip area corresponding to the sub-image, It is the ideal thickness for wafer polishing. is the average value of the strong edge curvature of all bright stripes in the sub-image, is the average value of the weak edge curvature of all bright stripes in the sub-image, is the average gradient of all pixels with strong edges in the sub-image, is the average gradient value of the pixels of all weak edges in the sub-image; the curvature is positively correlated with the sum of the differences of the horizontal coordinates of the pixels of each edge; the polishing state of the wafer is detected according to the polishing accuracy.

[0007] The effect is: by analyzing the local thickness distribution differences through machine vision and image processing methods, accurate detection of the polishing status of the wafer is achieved, and the polishing parameters are adjusted according to the detection results.

[0008] Furthermore, the polishing state of the wafer is detected according to the polishing accuracy, including: determining the polishing action according to the polishing accuracy; increasing the polishing pressure and rotation speed in response to the polishing accuracy of the sub-image being greater than 0; stopping polishing the wafer area corresponding to the sub-image in response to the polishing accuracy of the sub-image being less than or equal to 0, and issuing an over-polishing warning according to the absolute value of the polishing accuracy.

[0009] The effect is that the polishing pressure and rotation speed can be adjusted according to the current polishing precision to improve the polishing effect.

[0010] Furthermore, an over-polishing warning is issued based on the absolute value of the polishing accuracy, including: in response to the absolute value of the polishing accuracy of the sub-image being greater than a first threshold, it is determined that the wafer polishing does not meet the standard, and an alarm is issued; in response to the absolute value of the polishing accuracy of the sub-image being less than or equal to the first threshold, it is determined that the wafer polishing meets the standard.

[0011] Furthermore, a strong edge bright stripe region is obtained based on an adaptive region growing method, including: for any weak edge bright stripe region in a sub-image, a peak point in the region is selected as a region growing point, and growing is performed with a first growth threshold, wherein the first growth threshold is positively correlated with the roughness of the sub-image and the standard deviation of the grayscale value of the region; in response to the absence of a pixel point that meets the first growth threshold in the region, the growth is stopped and a first growth region is obtained; the threshold is updated in the first growth region and the growth is repeated until no new pixel points are added to the generated region after the threshold is updated, thereby obtaining a strong edge bright stripe region.

[0012] The effect is that the strong-edge bright stripe area is obtained by the adaptive region growing method, which avoids the situation that when the traditional binary segmentation interference image is used, the edge features of the divided interference stripe area will be weak edges, resulting in the inability to accurately obtain the stripe area features and then misjudge the polishing state of the chip.

[0013] Furthermore, updating the threshold in the first growth area and growing again includes: the updating threshold includes: selecting an array centered on the peak point and completely containing the first growth area as the initial array, calculating a second growth threshold, and growing with the second growth threshold, wherein the second growth threshold is positively correlated with the standard deviation of all grayscale values ​​within the eight neighborhoods of the initial array, and is also positively correlated with the roughness; in response to the absence of a pixel point that meets the second growth threshold in the first growth area, stopping the growth and obtaining the second growth area.

[0014] Furthermore, the calculated thickness of the chip area is corrected, including: constructing a data set based on four features of chip number, actual thickness, measured thickness, and roughness in historical data, wherein the measured thickness is the calculated thickness of the chip area; and constructing a nonlinear formula between the actual thickness, measured thickness, and roughness, obtaining parameters in the nonlinear formula based on a minimization objective function, correcting the measured thickness according to the parameters in the nonlinear formula to obtain a corrected chip thickness, and calculating the polishing accuracy based on the corrected chip thickness.

[0015] The effect is that the currently calculated thickness is corrected through historical data, thereby improving the accuracy of the measured thickness.

[0016] Furthermore, the bright stripe area in the sub-image is calculated The degree of strong right edge curvature , the calculation formula is: ; In the formula, is the number of pixels on the strong right edge, For the strong right edge The horizontal coordinate value of the pixel point, For the strong right edge The horizontal coordinate value of the pixel point.

[0017] The effect is that the polishing state of the wafer is characterized by the edge curvature degree, thereby improving the accuracy of subsequent calculation of the wafer polishing precision.

[0018] Further, calculate the second growth threshold , the calculation formula is: ; In the formula, is the standard deviation of all gray values ​​in the eight neighborhoods of the initial array, is the roughness of the sub-image, is the standard deviation of all gray values ​​in the first growth area of ​​the sub-image.

[0019] Furthermore, the pixel point that meets the second growth threshold includes: in response to the grayscale value of the pixel point being within the range from the grayscale value of the peak point to the second growth threshold, the pixel point meets the second growth threshold.

[0020] Beneficial effects:

[0021] (I) The present invention divides the interference image of the chip into multiple sub-images of equal size, realizes adaptive growth area segmentation of the sub-image based on the roughness of each sub-image to extract strong edge features, analyzes the thickness distribution difference according to the strong and weak edge features of the extracted interference bright fringes, calculates the polishing accuracy of the chip, and realizes accurate detection of the polishing state.

[0022] (ii) The currently calculated thickness is corrected using historical data to avoid the problem of deviation in thickness calculation caused by the different effects of roughness in different areas on the optical path and refractive index. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The following detailed description is read with reference to the accompanying drawings, which illustrate several embodiments of the present invention in an exemplary and non-limiting manner, and in which like or corresponding reference numerals represent like or corresponding parts, wherein:

[0024] Figure 1 is a flow chart schematically illustrating calculation of wafer polishing accuracy according to an embodiment of the present invention;

[0025] Figure 2 is a schematic diagram schematically illustrating an initial array and its eight neighborhoods according to an embodiment of the present invention;

[0026] Figure 3 FIG. 4 is a schematic diagram schematically illustrating the extraction of a strong edge bright stripe region according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0028] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0029] In one embodiment, the present embodiment divides the interference image of the wafer into multiple sub-images of equal size, realizes adaptive growth region segmentation of the sub-image based on the roughness of each sub-image to extract strong edge features, analyzes thickness distribution differences based on the extracted strong and weak edge features of the interference bright fringes to calculate the wafer polishing accuracy, and realizes accurate detection of the polishing state. It also includes correcting the measured thickness to obtain the actual thickness, thereby improving the accuracy of the detection of the polishing state.

[0030] like Figure 1 As shown, S101: acquiring an interference image of the wafer.

[0031] In one embodiment, an interferometer is used to measure the wafer, and the interference fringes and interference images of the wafer before and during polishing are recorded, and the thickness of different parts of the wafer is analyzed through the interference fringes and interference images of the wafer. In addition, a spectrometer is used to collect the reflection spectrum data of the wafer for subsequent analysis of the roughness of the wafer.

[0032] S102: Divide the interference image into equal parts to obtain multiple sub-images.

[0033] In one embodiment, during the detection of wafer polishing, due to the influence of factors such as wafer thickness, roughness, polishing disk size, polishing pressure, slurry flow rate, rotation speed, etc., the polishing speed and degree of different regions of the wafer are different, which makes it difficult to adjust the polishing parameters according to the local polishing state when analyzing the interference image of the entire wafer, and then the situation of excessive polishing is uncontrollable. Therefore, in this embodiment, the interference image is divided into equal parts to obtain multiple sub-images, and the roughness of the corresponding wafer area and the thickness of the corresponding wafer area in each sub-image are calculated, so as to perform polishing analysis. It should be noted that calculating the roughness of the wafer and the thickness of the corresponding wafer area is a method of the prior art and will not be described in detail here.

[0034] S103: Calculate the polishing accuracy of the sub-image.

[0035] In one embodiment, during the detection of the wafer polishing state, because the wafer polishing state detection is affected by factors such as polishing uniformity and thickness, when a single thickness variation feature is used to judge the real-time wafer polishing state, it may be misjudged that the current polishing state has uneven polishing or over-polishing in some areas. In this case, according to the edge features of the interference fringe image, the current interference fringe spacing variation and the interference fringe displacement deviation difference are calculated to analyze the thickness distribution difference, and the wafer polishing accuracy is calculated in combination with the thickness variation, so that the current wafer polishing state can be accurately analyzed.

[0036] In one embodiment, to calculate the polishing accuracy of the sub-image, it is necessary to first obtain a grayscale image of the sub-image, and perform binary segmentation on the grayscale image to obtain a weak edge bright stripe area ( Figure 3 The dotted area in the figure), where the binary segmentation method can use the Otsu threshold segmentation method. However, the edge of the weak edge bright stripe area is not clear enough. In order to improve the subsequent calculation results, the strong edge bright stripe area is obtained by the adaptive region growing method ( Figure 3 ).

[0037] In one embodiment, a strong edge bright stripe region is obtained by an adaptive region growing method, including: for any weak edge bright stripe region in a sub-image, a peak point in the weak edge bright stripe region is selected as a region growing point, and growth is performed with a first growth threshold, wherein the first growth threshold is positively correlated with the roughness of the sub-image and the standard deviation of the gray value of the region, and the first growth threshold is equal to , is the roughness of the sub-image, The standard deviation of all gray values ​​in the weak edge bright stripe area, is the rounding down symbol; if there are no pixels that meet the first growth threshold in the weak edge bright stripe region, the growth is stopped and the first growth region is obtained. Among them, the pixels that meet the first growth threshold refer to the grayscale values ​​of the pixels in the strong edge bright stripe region that are within the range from the grayscale value of the peak point to the second growth threshold. Then, the threshold is updated in the first growth region and the growth is repeated until no new pixels are added to the generation region after the threshold is updated, and the strong edge bright stripe region is obtained.

[0038] In one embodiment, updating the threshold in the first growing region and growing again comprises: the updating threshold comprises: selecting an array centered at the peak point and completely containing the first growing region as the initial array ( Figure 2 The shaded area in the figure is the initial array), the second growth threshold is calculated, and the growth is performed with the second growth threshold, and the second growth threshold is consistent with the eight neighborhoods of the initial array ( Figure 2The dotted area in the figure is an eight-neighborhood) is positively correlated with the standard deviation of all grayscale values ​​within the first growth area, and is also positively correlated with the roughness; in response to the absence of pixels that meet the second growth threshold in the first growth area, the growth is stopped and the second growth area is obtained.

[0039] In one embodiment, the second growth threshold is calculated , the calculation formula is: ; In the formula, is the standard deviation of all gray values ​​in the eight neighborhoods of the initial array, is the roughness of the sub-image, is the standard deviation of all gray values ​​in the first growth area of ​​the sub-image, The floor symbol.

[0040] Exemplarily, the first growth includes: obtaining a first growth area after passing the first growth threshold; the second growth includes: selecting an array in the first growth area centered on the peak point and completely containing the first growth area as the initial array, and growing with the second growth threshold, if there is a pixel point that meets the second growth threshold, the pixel point is added to the second growth area; then the third growth is performed, the third growth includes: selecting an array in the first growth area centered on the peak point and completely containing the second growth area as the initial array, and growing with the third growth threshold, the third growth threshold is calculated in the same way as the second growth threshold, until no new pixels are added during a certain growth, then the growth stops, and a strong edge bright stripe area is obtained.

[0041] In one embodiment, the polishing accuracy of the sub-image is calculated , the calculation formula is: ; In the formula, is the thickness of the chip area corresponding to the sub-image, It is the ideal thickness for wafer polishing. is the average value of the strong edge curvature of all bright stripes in the sub-image, is the average value of the weak edge curvature of all bright stripes in the sub-image, is the average gradient of all pixels with strong edges in the sub-image, is the average gradient of all weak edge pixels in the sub-image. The larger the value, the more uneven the thickness of the wafer is or the more defects there are.

[0042] In one embodiment, the curvature of each edge in the sub-image is calculated. Specifically, all the interference bright fringe areas in the sub-image are counted (eg Figure 3 As shown), we get , Indicates bright stripe areas, calculate the strong right edge curvature or strong left edge curvature of all bright stripe areas in the sub-image, and the calculation formula is: ; In the formula, Indicates the bright stripe area in the sub-image The degree of strong right edge curvature or strong left edge curvature, is the number of pixels on the strong right edge or strong left edge, The first The horizontal coordinate value of the pixel point, The first The horizontal coordinate value of the pixel point is calculated, and the average value of the strong right edge curvature and the strong left edge curvature is taken as the strong edge curvature; it should be noted that the bright stripe weak left edge curvature, the bright stripe weak right edge curvature and the weak edge curvature are calculated in the same way as above, which will not be described in detail here.

[0043] For example, taking the strong right edge of the bright stripe as an example, calculating the curvature of the strong right edge of the bright stripe includes: counting the bright stripe area The coordinate set of the medium strong right edge pixel points is arranged in the order of the strong right edge pixels from bottom to top. In the process of calculating the curvature, the coordinates are also subtracted from bottom to top; then the bright stripe area is calculated The degree of strong right edge curvature , the calculation formula is: In the formula, is the number of pixels on the strong right edge, For the strong right edge The horizontal coordinate value of the pixel point, For the strong right edge The horizontal coordinate value of the pixel point is calculated, and the average value of the strong left edge curvature and the strong right edge curvature is taken as the strong edge curvature.

[0044] In another embodiment, it should be noted that in the process of calculating the thickness of the wafer area corresponding to different sub-images, because the roughness of different areas has different effects on the optical path and refractive index, the calculated thickness is biased and different from the actual thickness, so the polishing degree of the wafer may be misjudged, resulting in over-polishing or under-polishing. Therefore, in order to improve the accuracy of the detection of the polishing state, in this embodiment, the currently calculated thickness can also be corrected in combination with historical data.

[0045] In one embodiment, the correction process includes: according to the wafer number, actual thickness in the historical data , measure thickness , Roughness Four features are used to construct a data set, wherein the measured thickness is the calculated thickness of the wafer area; and a nonlinear formula between the actual thickness, the measured thickness and the roughness is constructed, and the nonlinear formula is: Where and is the parameter to be fitted, for to Integers between . Get the parameters in the nonlinear formula based on minimizing the objective function and , the measured thickness is corrected according to the parameters in the nonlinear formula to obtain the corrected thickness of the wafer, and the polishing accuracy is calculated based on the corrected thickness of the wafer.

[0046] In one embodiment, the objective function is: ;in It is Measure thickness of each wafer, It is The actual thickness of the wafer, It is The roughness of the wafer, for to Integers between and is the parameter to be fitted. Then use the gradient descent method to minimize the objective function, repeatedly calculate the gradient and update the parameters until the maximum number of iterations is reached. Stop the iteration when and value.

[0047] S104: Detecting the polishing state of the wafer according to the polishing accuracy.

[0048] In one embodiment, the polishing action is determined according to the polishing accuracy; if the polishing accuracy of the sub-image is greater than 0, it means that the thickness of the wafer is still greater than the ideal thickness, and the polishing pressure and rotation speed are appropriately increased at this time; if the polishing accuracy of the sub-image is less than or equal to 0, the polishing of the wafer area corresponding to the sub-image is stopped, and an over-polishing warning is issued according to the absolute value of the polishing accuracy, wherein, when the polishing accuracy of the sub-image is equal to 0, it means that the thickness of the wafer has reached the ideal thickness, and polishing is stopped at this time; when the polishing accuracy of the sub-image is less than 0, it means that the thickness of the wafer is lower than the ideal thickness, and an over-polishing warning is issued according to the absolute value of the polishing accuracy.

[0049] In one embodiment, an over-polishing warning is performed according to the absolute value of the polishing precision, including: if the absolute value of the polishing precision of the sub-image is greater than a first threshold, it is determined that the wafer polishing does not meet the standard, and an alarm is issued; if the absolute value of the polishing precision of the sub-image is less than or equal to the first threshold, it is determined that the wafer polishing meets the standard. In this embodiment, the empirical value of the first threshold is 0.2, and in other embodiments, the empirical value of the first threshold is 0.1 or 0.15, etc., which can be set according to specific actual conditions.

[0050] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0051] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0052] The above-mentioned embodiments only express several implementation methods of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent application. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A wafer polishing state detection method based on machine vision, characterized in that: include: acquiring an interference image of the wafer; Divide the interference image into equal parts to obtain a plurality of sub-images, and calculate the roughness of the wafer area and the thickness of the wafer area corresponding to each sub-image; Obtain a grayscale image of the sub-image, perform binary segmentation on the grayscale image, obtain a weak edge bright stripe region, and obtain a strong edge bright stripe region based on an adaptive region growing method, and calculate the polishing accuracy of the sub-image , the calculation formula is: ； In the formula, is the thickness of the chip area corresponding to the sub-image, It is the ideal thickness for wafer polishing. is the average value of the strong edge curvature of all bright stripes in the sub-image, is the average value of the weak edge curvature of all bright stripes in the sub-image, is the average gradient of all pixels with strong edges in the sub-image, is the average gradient of all pixels of weak edges in the sub-image, where the degree of curvature is positively correlated with the sum of the differences in the horizontal coordinates of the pixels of the edge; The polishing state of the wafer is detected according to the polishing accuracy.

2. The wafer polishing state detection method based on machine vision according to claim 1, characterized in that: According to the polishing accuracy, the polishing state of the wafer is detected, including: Determining the polishing action according to the polishing accuracy; In response to the polishing accuracy of the sub-image being greater than 0, increasing the polishing pressure and the rotation speed; In response to the polishing accuracy of the sub-image being less than or equal to 0, polishing of the wafer region corresponding to the sub-image is stopped, and an over-polishing warning is issued according to the absolute value of the polishing accuracy.

3. The method for detecting wafer polishing state based on machine vision according to claim 2, characterized in that: According to the absolute value of polishing accuracy, over-polishing warning is issued, including: In response to the absolute value of the polishing accuracy of the sub-image being greater than a first threshold, it is determined that the wafer polishing does not meet the standard, and an alarm is issued; In response to the absolute value of the polishing accuracy of the sub-image being less than or equal to a first threshold, it is determined that the wafer polishing meets the standard.

4. The wafer polishing state detection method based on machine vision according to claim 1, characterized in that: Based on the adaptive region growing method, the strong edge bright stripe region is obtained, including: For any weak edge bright stripe region in the sub-image, a peak point in the region is selected as a region growth point, and growth is performed with a first growth threshold, wherein the first growth threshold is positively correlated with the roughness of the sub-image and the standard deviation of the grayscale value of the region; In response to the absence of a pixel point meeting the first growth threshold in the region, stopping the growth and obtaining the first growth region; In the first growing area, the threshold is updated and the area is grown again until no new pixel points are added to the generating area after the threshold is updated, thereby obtaining a bright stripe area with a strong edge.

5. The method for detecting wafer polishing state based on machine vision according to claim 4, characterized in that: Updating the threshold in the first growing area and growing again includes: The updating threshold comprises: selecting an array centered at the peak point and completely containing the first growth area as the initial array, calculating a second growth threshold, and growing with the second growth threshold, wherein the second growth threshold is positively correlated with the standard deviation of all gray values ​​in the eight neighborhoods of the initial array and is also positively correlated with the roughness; In response to the fact that there is no pixel point meeting the second growth threshold in the first growth area, the growth is stopped and the second growth area is obtained.

6. The wafer polishing state detection method based on machine vision according to claim 1, characterized in that: Corrections are made to the calculated thickness of the wafer area, including: Constructing a data set based on four features of wafer number, actual thickness, measured thickness, and roughness in historical data, wherein the measured thickness is the thickness of the wafer area calculated; A nonlinear formula between the actual thickness, measured thickness and roughness is constructed, parameters in the nonlinear formula are obtained based on the minimization objective function, the measured thickness is corrected according to the parameters in the nonlinear formula to obtain the corrected thickness of the wafer, and the polishing accuracy is calculated based on the corrected thickness of the wafer.

7. The method for detecting wafer polishing state based on machine vision according to claim 1, characterized in that: Calculate the degree of strong edge curvature of the bright stripe area in the sub-image, including: The strong right edge curvature or the strong left edge curvature of the bright stripe region in the sub-image is calculated, and the calculation formula is: ； In the formula, Indicates the bright stripe area in the sub-image The degree of strong right edge curvature or strong left edge curvature, is the number of pixels on the strong right edge or strong left edge, The first The horizontal coordinate value of the pixel point, The first The horizontal coordinate value of the pixel point; The average value of the strong right edge curvature degree and the strong left edge curvature degree is taken as the strong edge curvature degree.

8. The method for detecting wafer polishing state based on machine vision according to claim 5, characterized in that: Calculate the second growth threshold , the calculation formula is: ； In the formula, is the standard deviation of all gray values ​​in the eight neighborhoods of the initial array, is the roughness of the sub-image, is the standard deviation of all gray values ​​in the first growth area of ​​the sub-image.

9. The method for detecting wafer polishing state based on machine vision according to claim 5, characterized in that: Pixels that meet the second growth threshold include: In response to the grayscale value of the pixel being within the range from the grayscale value of the peak point to the second growth threshold, the pixel meets the second growth threshold.

Images