Image detection method for particle pollution on surface of semiconductor wafer

Abstract

The invention relates to the technical field of semiconductors, in particular to an image detection method for particle pollution on the surface of a semiconductor wafer. The method comprises the steps of obtaining a global noise characterization value according to a pixel value variance in a local window and an area proportion of a first target connected domain on a binary image, obtaining a region fuzzy characterization value according to the edge total length of all broken connected domains, the edge total length of all second target connected domains and a cavity area of the second target connected domains, and obtaining a region fuzzy characterization value according to the region fuzzy characterization value. Obtaining a morphological processing strategy corresponding to the surface suppression image according to the global noise characterization value and the region fuzzy characterization value, and performing morphological processing on the surface suppression image according to the morphological processing strategy to obtain a pollution particle to-be-detected image. And carrying out pollution particle detection and identification on the to-be-detected image of the pollution particles by utilizing a pollution particle detection model. And the accuracy and the integrity of detecting and identifying the pollution particles on the surface of the semiconductor wafer can be improved.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor technology, and particularly relates to an image detection method for particle contamination on the surface of a semiconductor wafer. Background Art

[0002] A semiconductor wafer is the basic material for manufacturing semiconductor chips (such as integrated circuits, microprocessors, memories, etc.), and semiconductor chips are the core components of the modern electronics industry and are widely used in fields such as computers, communication devices, consumer electronics, and automotive electronics; during the production process of semiconductor wafers, air flow disturbances, equipment operation, temperature and humidity fluctuations, etc. can all cause contamination particles to appear on the surface of the semiconductor wafer. For example, air flow disturbances or equipment operation may still introduce particles such as silicon chips and metal oxides in the air. Too low humidity is likely to generate electrostatic adsorption of particles, and too high humidity may condense aerosols to form contamination particles. The contamination particles that appear on the surface of the semiconductor wafer will seriously affect the application performance of the manufactured semiconductor chips. Therefore, it is crucial to identify and detect the contamination particles on the surface of the semiconductor wafer currently.

[0003] Moreover, since the surface of a semiconductor wafer usually has complex circuit patterns or lithography structures, and currently, in order to avoid the interference and influence of complex circuit patterns or lithography structures on the detection of contamination particles on the wafer surface, generally, a Fourier transform is performed on the collected image of the semiconductor wafer surface to obtain a frequency-domain image. In order to remove the interference of circuit patterns or lithography structures, usually, a low-pass filter is used to filter the obtained spectral image, and an inverse Fourier transform is performed on the filtered image. Finally, an existing contamination particle detection model is used to detect and identify the contamination particles in the image obtained by the inverse Fourier transform. However, performing low-pass filtering on the obtained frequency-domain image will cause the characteristics of the contamination particles to become blurred. If the cut-off frequency is set too low during low-pass filtering, the edges of the contamination particles will be overly suppressed, resulting in the blurring of the contamination particle characteristics. This blurring of the contamination particle characteristics will directly have a negative impact on the subsequent detection and identification of contamination particles, that is, it will cause problems of inaccurate and incomplete identification when identifying the contamination particle region subsequently. Therefore, how to improve the accuracy and integrity of detecting and identifying the contamination particles on the surface of a semiconductor wafer has become an urgent problem to be solved. Summary of the Invention

[0004] In order to solve the above problems, the present invention provides an image detection method for particle contamination on the surface of a semiconductor wafer, and the specific technical solution adopted is as follows:

[0005] An embodiment of the present invention provides an image detection method for particle contamination on the surface of a semiconductor wafer, including the following steps:

[0006] Acquiring a surface suppression image of a semiconductor wafer to be inspected;

[0007] The surface suppression image is divided into all local windows on the surface suppression image using a preset window, a first target connected domain and a second target connected domain on the binary image of the surface suppression image are obtained, and a global noise representation value is obtained according to the variance of pixel values in the local windows and the area ratio of the first target connected domain on the binary image; broken connected domains in all second target connected domains are obtained, and a regional fuzzy representation value is obtained according to the total edge lengths of all broken connected domains, the total edge lengths of all second target connected domains, and the hole area of the second target connected domain;

[0008] According to the global noise characterization value and the regional fuzzy characterization value, a morphological processing strategy corresponding to the surface suppression image is obtained, and the surface suppression image is morphologically processed according to the morphological processing strategy to obtain an image of contaminated particles to be detected, and the contaminated particles are detected on the image of contaminated particles to be detected using a contaminated particle detection model.

[0009] Beneficial effect: The present invention first obtains a surface suppression image of a semiconductor wafer to be inspected; then uses a preset window to divide the surface suppression image to obtain all local windows on the surface suppression image, obtains the first target connected domain and the second target connected domain on the binary image of the surface suppression image, and obtains a global noise characterization value based on the pixel value variance in the local window and the area ratio of the first target connected domain on the binary image; then obtains the broken connected domain in all the second target connected domains, and obtains a regional fuzzy characterization value based on the total edge length of all the broken connected domains, the total edge length of all the second target connected domains and the hole area of the second target connected domain; finally, according to the global noise characterization value and the regional fuzzy characterization value, obtains a morphological processing strategy corresponding to the surface suppression image, and performs morphological processing on the surface suppression image according to the morphological processing strategy to obtain a contaminated particle image to be detected, and uses a contaminated particle detection model to detect and identify contaminated particles on the contaminated particle image to be detected. The present invention performs morphological processing on the surface suppression image, that is, by repairing the fuzzy features of the surface suppression image, it can minimize the impact of the fuzzy phenomenon of the contamination particle features on the accuracy and integrity of contamination particle identification, thereby improving the accuracy and integrity of the detection and identification of contamination particles on the surface of the semiconductor wafer. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. 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 any creative work.

[0011] Figure 1 This is a flow chart of an image detection method for particle contamination on the surface of a semiconductor wafer according to the present invention. DETAILED DESCRIPTION

[0012] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field fall within the scope of protection of the embodiments of the present invention.

[0013] 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 pertains.

[0014] This embodiment provides an image detection method for particle contamination on the surface of a semiconductor wafer, which is described in detail as follows:

[0015] like Figure 1 As shown, the image detection method for particle contamination on the surface of a semiconductor wafer includes the following steps:

[0016] Step S001 : obtaining a surface suppression image of a semiconductor wafer to be inspected.

[0017] This embodiment mainly uses morphology to repair the image obtained by inverse transformation, or mainly uses morphology to repair the image after removing the interference of circuit patterns or photolithography structures, thereby reducing the influence of interference such as circuit patterns or photolithography structures on the accuracy and integrity of contamination particle identification, while also reducing the influence of fuzzy contamination particle features on the accuracy and integrity of contamination particle identification as much as possible, thereby improving the accuracy and integrity of detection and identification of contamination particles on the surface of semiconductor wafers. That is, this embodiment can not only reduce the influence of circuit patterns or photolithography structures on the surface of semiconductor wafers on the accuracy and integrity of detection and identification of contamination particles on the surface of semiconductor wafers as much as possible, but also reduce the influence of fuzzy contamination particle features on the accuracy and integrity of detection and identification of contamination particles on the surface of semiconductor wafers as much as possible, thereby achieving the purpose of improving the accuracy and integrity of detection and identification of contamination particles on the surface of semiconductor wafers.

[0018] Since all the methods for detecting and identifying surface contamination particles on semiconductor wafers are the same in this embodiment, for the sake of easy understanding, the following will analyze the detection and identification process of surface contamination particles on any semiconductor wafer to be detected as an example. From the above description, it can be seen that this embodiment mainly uses morphology to repair the image after removing the interference of circuit patterns or lithography structures, so as to improve the accuracy and integrity of detecting and identifying surface contamination particles on semiconductor wafers. Therefore, this embodiment first needs to obtain the image after removing the interference of circuit patterns or lithography structures, that is, the surface suppression image of the semiconductor wafer to be detected. The specific obtaining process is as follows:

[0019] First, use high-resolution optical microscopes, CCD / CMOS cameras and other devices to scan the surface of the semiconductor wafer, and record the high-quality image obtained by the scan as the surface original image of the semiconductor wafer to be detected. And there are no obvious acquisition problems in the collected surface original image. The acquisition problems include but are not limited to insufficient exposure, out-of-focus, etc.; then perform image preprocessing on the collected surface original image, and record the surface original image after image preprocessing as the initial processed image. And the image preprocessing includes but is not limited to gray-scale stretching or histogram equalization, etc. Gray-scale stretching or histogram equalization can improve the contrast of the image and highlight the detailed features on the wafer surface for subsequent processing.

[0020] Since the surface of a semiconductor wafer usually has complex circuit patterns or lithography structures, the circuit patterns on the surface of the semiconductor wafer usually exhibit periodicity or regularity. Currently, in order to avoid the interference and influence of complex circuit patterns or lithography structures on the detection of contamination particles on the wafer surface, it is necessary to perform a Fourier transform on the obtained initial processed image, that is, use the Fourier transform to perform a frequency-domain conversion on the initial processed image, and record the image obtained by the conversion as the frequency-domain image. The high-frequency information part on the frequency-domain image obtained by the Fourier transform usually corresponds to the edge and detail information of the wafer, and the low-frequency information part usually corresponds to the circuit pattern or lithography structure information. Moreover, the circuit patterns on the surface of the semiconductor wafer usually exhibit periodicity or regularity. Therefore, in the frequency domain, the circuit patterns on the surface of the semiconductor wafer are represented as low-frequency components or concentrated in the central region of the spectrum, while contamination particles or noise are mostly minute features corresponding to high-frequency components. It can also be said that the low-frequency components contain the overall gray-scale distribution and approximate texture pattern of the wafer, and the high-frequency components are distributed in the edge region of the spectrum. In order to remove the interference and influence of the periodic circuit patterns or lithography structures on the surface of the wafer on the subsequent detection of contamination particles on the wafer surface, a low-pass filter is first used to filter the obtained frequency-domain image, and the image obtained after filtering is recorded as the filtered frequency-domain image. Then, an inverse Fourier transform is performed on the obtained filtered frequency-domain image, and the image obtained by the inverse transform is recorded as the surface suppression image of the semiconductor wafer to be detected. Moreover, performing a low-pass filter after performing a Fourier transform on the initial processed image and then performing an inverse Fourier transform can remove the low-frequency information on the image, that is, can remove the circuit pattern or lithography structure information on the image as much as possible, thereby reducing the interference and influence of complex circuit patterns or lithography structures on the subsequent detection of contamination particles on the wafer surface. That is, the surface suppression image is an image after removing or suppressing the circuit pattern or lithography structure information.

[0021] Therefore, through the above process, the surface suppression image of the semiconductor wafer to be detected can be obtained in this embodiment.

[0022] Step S002: Divide the surface suppression image using a preset window to obtain all local windows on the surface suppression image, obtain the first target connected domain and the second target connected domain on the binary image of the surface suppression image, and obtain a global noise characterization value according to the pixel value variance in the local window and the area ratio of the first target connected domain on the binary image; obtain the broken connected domains in all the second target connected domains, and obtain a regional blur characterization value according to the total edge length of all the broken connected domains, the total edge length of all the second target connected domains, and the hole area of the second target connected domain.

[0023] When performing low-pass filtering on the acquired frequency-domain image, it will cause the characteristics of the contamination particles to become blurred. The phenomenon of the blurred characteristics of the contamination particles caused by the low-pass filtering process will also result in the blurred characteristics of the contamination particles on the surface suppression image obtained by performing the inverse Fourier transform. If the cut-off frequency is set too low during the low-pass filtering, it will cause the edges of the contamination particles to be overly suppressed, resulting in the blurred characteristics of the contamination particles, which will affect the subsequent detection and recognition of the contamination particles. In addition, spectral leakage will also occur in the Fourier transform. The spectral leakage that occurs in the Fourier transform may also cause the spectral energy of the circuit pattern to leak into the high-frequency region, ultimately resulting in the blurred characteristics of the contamination particles and seriously interfering with the subsequent detection and recognition of the contamination particles. Therefore, when there is a phenomenon of blurred contamination particle characteristics on the surface suppression image, if the surface suppression image is directly used to identify the contamination particles on the surface of the semiconductor wafer to be detected subsequently, problems such as incomplete and inaccurate identification of the contamination particle area will occur. That is, if the surface suppression image is directly input into the contamination particle detection model, it will cause problems of inaccurate and incomplete identification of the contamination particle area. In order to avoid the negative impact of the blurred contamination particle characteristics on the contamination particle recognition as much as possible in this embodiment, the present embodiment will next use morphology to repair the obtained surface suppression image. That is to say, the present embodiment will next use morphology to repair the blurred contamination particle characteristics on the surface suppression image, and then use the repaired image to detect and recognize the contamination particles subsequently, so as to achieve the purpose of improving the accuracy and integrity of the detection and recognition of the contamination particles on the surface of the semiconductor wafer. And in order to further ensure the accuracy and integrity of the subsequent contamination particle detection and recognition, noise removal will be further performed during the repair process.

[0024] In addition, since morphology includes two operations, i.e., opening operation and closing operation, and the purposes achieved by these two operations and the sequence of these two operations have a great impact on the effect of restoration and noise removal. If an inappropriate operation method or operation sequence is selected, not only the purpose of restoring the characteristic information of contaminated particles cannot be achieved, but also image information may be further lost or new artifacts may be generated. Therefore, only by selecting the correct operation method or operation sequence can the blurred contaminated particle characteristics be restored and the noise be removed. For example, when there is a lot of noise in the image, the opening operation should be performed first to remove the noise, and then the closing operation should be performed to restore the blurred contaminated particle characteristics, so as to avoid noise interfering with the subsequent closing operation restoration. When the edges of the contaminated particles are severely broken, that is, when the degree of blurring of the contaminated particle characteristics is relatively high, the closing operation should be performed first to restore, and then the opening operation should be performed to remove the noise, so as to avoid the opening operation further weakening the contaminated particle characteristics. In addition, it should be noted that the opening operation first erodes and then dilates, mainly used to remove small objects and smooth the boundaries of larger objects. The closing operation first dilates and then erodes, mainly used to fill small holes, bridge small cracks and smooth the boundaries of objects. Therefore, the erosion operation of the opening operation will remove small objects and protruding parts on the boundary in the image, and the subsequent dilation operation of the opening operation will restore the shape of larger objects to a certain extent, but will not restore the small objects that have been completely eroded. That is, the opening operation can remove small noise points and isolated small areas in the image and keep the shape and position of larger objects basically unchanged. The dilation operation of the closing operation will expand the boundaries of objects and fill the small hole part, and the subsequent erosion operation of the closing operation will shrink the expanded boundary to make the object return to a size close to the original size, but at this time the small holes and cracks have been filled. That is, the closing operation can fill small holes and bridge small cracks in the image and keep the overall shape of the object. At this time, the object mainly refers to the contaminated particles.

[0025] Therefore, based on the above analysis, it can be seen that the noise level and the degree of blurring of the contaminated particles shown in the surface suppression image are the basis for determining the type of morphological operation and the sequence of morphological operations. Therefore, in the following, the noise level and the degree of blurring of the contaminated particles need to be analyzed. That is to say, in the following, the global noise characterization value and the regional blurring characterization value need to be obtained. The global noise characterization value and the regional blurring characterization value can respectively characterize the noise level and the degree of blurring of the contaminated particles. Then the specific acquisition process of the global noise characterization value and the regional blurring characterization value is as follows:

[0026] First, obtain a preset window and place the preset window at the top-left corner of the surface suppression image. Then, slide it from left to right and top to bottom. All the windows obtained after the sliding on the surface suppression image are recorded as the local windows on the surface suppression image. The purpose of dividing the surface suppression image to obtain local windows is to analyze the amount of noise in the image later. Sliding from left to right and top to bottom means moving a preset step length to the right each time. After completing a row scan, the window moves down a preset step length to the next row and starts sliding from the left again until all the pixel points on the surface suppression image are traversed. Moreover, when the right boundary of the window exceeds the image width, the sliding stops and jumps to the starting position of the next row. When the window exceeds the image boundary, the window size will also be automatically adjusted to retain the valid area. Additionally, in a specific application, the implementer needs to set the size of the preset window and the size of the preset step length according to the actual situation such as the size of the surface suppression image. However, in this embodiment, it is required that the preset window is a rectangular window and the preset step length is the side length of the preset window. For example, in this embodiment, the size of the preset window can be set to 32×32 or 64×64. Then, when the size of the preset window is 32×32, the corresponding preset step length is 32. When the size of the preset window is 64×64, the corresponding preset step length is 64.

[0027] After obtaining the local windows, the Otsu algorithm is used to perform binary processing on the surface suppression image to obtain the binary image of the surface suppression image. Then, connected component analysis is performed on the obtained binary image, and all the connected components obtained from the connected component analysis are recorded as the connected components to be analyzed on the binary image. Then, the area of each connected component to be analyzed is obtained, and the area of the connected component to be analyzed refers to the number of pixel points in the corresponding connected component to be analyzed. After that, based on the area of the connected component to be analyzed and the preset connected component area threshold, the first target connected component and the second target connected component are obtained. The first target connected component and the second target connected component are an important basis for obtaining the global noise characterization value and the regional blur characterization value later. Moreover, the probability that the area of the first target connected component is small and it is noise is relatively large, and the probability that the area of the second target connected component is large and it is a contaminated particle region is relatively large. The specific obtaining process of the first target connected component and the second target connected component is as follows:

[0028] For any connected region to be analyzed, it is determined whether the area of the connected region to be analyzed is less than a preset connected region area threshold. If so, it indicates that the probability that the connected region to be analyzed is a noise region is relatively high. Therefore, the connected region to be analyzed is denoted as the first target connected region. Otherwise, it indicates that the probability that the connected region to be analyzed is a pollution particle region is relatively high. Therefore, the connected region to be analyzed is denoted as the second target connected region. And in specific applications, the implementer needs to set the preset connected region area threshold according to actual situations such as noise characteristics. For example, since generally the area of noise in an image will not be greater than 10, and salt-and-pepper noise usually appears as single isolated pixel points, in this embodiment, the preset connected region area threshold is set to 10.

[0029] After obtaining the local window and the first target connected region, the global noise characterization value is obtained according to the variance of the pixel values in the local window and the area ratio of the first target connected region on the binary image of the surface suppression image. The specific process of obtaining the global noise characterization value is as follows:

[0030] First, the variance of the pixel values of all pixel points in each local window is obtained, and the variance of the pixel values of all pixel points in each local window is normalized using the normalization function Norm(). The normalization result is denoted as the normalized variance corresponding to the local window. The value range of the normalized variance is from 0 to 1. Then, the number of local windows with a normalized variance greater than the preset variance threshold is counted and denoted as the quantity characterization value. The ratio of the quantity characterization value to the total number of local windows on the surface suppression image is calculated and denoted as the feature ratio. The feature ratio is normalized using the normalization function Norm(), and the normalization result is denoted as the first noise characterization value. And the larger the normalized variance corresponding to the local window, the more it indicates that the corresponding local window is a noise window. That is, in this embodiment, a local window with a normalized variance greater than the preset variance threshold is determined as a noise window. Therefore, the more the number of local windows with a normalized variance greater than the preset variance threshold, the more noise there is on the surface suppression image. In addition, in specific applications, the implementer needs to set the preset variance threshold according to actual situations such as accuracy requirements, experimental statistics, and the value range of the normalized variance. For example, in this embodiment, the preset variance threshold can be set to 0.5.

[0031] After obtaining the first noise characterization value, calculate the sum of the areas of all the first target connected regions on the binary image, and denote it as the first area characterization value. Then, calculate the ratio of the first area characterization value to the total number of pixel points in the binary image, and denote it as the proportion of the first target connected region area. Since the probability that the first target connected region is a noise connected region is relatively high, when the proportion of the first target connected region area is larger, it indicates that the noise distribution on the surface suppression image is stronger. Then, determine whether the proportion of the first target connected region area is less than a preset proportion threshold. If so, it indicates that the number of noise points on the surface suppression image is relatively small, or in other words, there may be only a few scattered noise points on the surface suppression image, and at this time, the influence of noise on the subsequent recognition of pollution particles on the surface suppression image is extremely small. Then, assign the second noise characterization value to 0, that is, when the proportion of the first target connected region area is less than the preset proportion threshold, the constant 0 is used as the second noise characterization value. When the proportion of the first target connected region area is greater than or equal to the preset proportion threshold, there may be some small noise clusters or a large amount of noise in the image, and at this time, the interference to the recognition of pollution particles is relatively large. Then, assign the second noise characterization value to 1, that is, when the proportion of the first target connected region area is greater than or equal to the preset proportion threshold, the constant 1 is used as the second noise characterization value, and the larger the proportion of the first target connected region area, the more noise there is in the image, and the greater the impact on the recognition of pollution particles. In addition, in specific applications, the implementer usually needs to preset the proportion threshold according to experimental statistics or actual situations. Since generally, when the noise proportion on the image reaches 10% of the overall image, it will cause interference or influence to some subsequent analyses of the image. For example, when the noise proportion on the image reaches 10% of the overall image, the possibility that key information on the image is covered by noise is greater. Therefore, in this embodiment, the preset proportion threshold is set to 10%.

[0032] After obtaining the first noise characteristic value and the second noise characteristic value, fuse the first noise characteristic value and the second noise characteristic value, and use the fusion result as the global noise characterization value. The larger the value of the global noise characterization value, the more noise there is on the surface suppression image, and vice versa, the smaller the value of the global noise characterization value, the less noise there is on the surface suppression image. Fusing the first noise characteristic value and the second noise characteristic value means performing a weighted sum on the first noise characteristic value and the second noise characteristic value, and denoting the weighted sum result as the global noise characterization value, that is, the global noise characterization value is , where is the first weight, R1 is the first noise characteristic value, R2 is the second noise characteristic value, is the second weight value; and in specific applications, the implementer needs to set the first weight value and the second weight value according to the actual situation or the importance of the noise characteristic values in different dimensions for the global noise characterization value of the image. If the first noise characteristic value and the second noise characteristic value in this embodiment are equally important for the global noise characterization value of the image, then both the first weight value and the second weight value are set to 0.5.

[0033] After obtaining the characterization value representing the degree or amount of noise on the surface suppression image, the second target connected region is analyzed, and the second target connected region is a suspected contaminated particle region. Based on the analysis result, the broken connected region in the second target connected region is obtained. The edge or contour length of the broken connected region is the key to subsequent analysis of the fuzziness of the contaminated particle characteristics. Then the specific process of obtaining the broken connected region is as follows:

[0034] For any second target connected region: First, use an edge detection algorithm to extract the edge of the second target connected region to obtain the edge of the second target connected region, and then determine whether the edge of the second target connected region is a non-closed edge. If so, it is determined that the edge of the second target connected region is a broken edge. Then at this time, the second target connected region is recorded as the broken connected region; in addition, a contour detection algorithm can also be used to detect the contour of the second target connected region, and then determine whether the detected contour is connected end to end or is a closed contour. If not, it is determined that the edge of the second target connected region is a broken edge. Then at this time, the second target connected region is recorded as the broken connected region; and it is well-known to perform contour detection or edge extraction on the connected region.

[0035] After obtaining the broken connected region, the hole area of the second target connected region is obtained. Obtaining the hole area is to analyze the proportion of the hole area, and the proportion of the hole area can reflect the fuzziness of the contaminated particle region. Then the specific process of obtaining the hole area of the second target connected region is as follows:

[0036] For any second target connected region, fill the second target connected region, and record the filled second target connected region as the filled connected region corresponding to the second target connected region. Calculate the result of subtracting the area of the filled connected region corresponding to the second target connected region from the area of the second target connected region, and record it as the hole area of the second target connected region. And the process of filling the connected region is well-known. For example, the contour filling method can be selected for filling.

[0037] After obtaining the hole area of the second target connected region, based on the total edge length of all broken connected regions, the total edge length of all second target connected regions, and the hole area of the second target connected region, a regional fuzziness characterization value is obtained. The size of the regional fuzziness characterization value can characterize the fuzziness of the contaminated particle region. Then the specific process of the regional fuzziness characterization value is as follows:

[0038] First, calculate the sum of the lengths of the edges of all broken connected regions on the binary image of the surface suppression image, and denote it as the broken edge length. Calculate the sum of the lengths of the edges of all second target connected regions on the binary image of the surface suppression image, and denote it as the total edge length. Calculate the ratio of the broken edge length to the total edge length, and denote it as the edge breakage rate. The length of the edge of any connected region refers to the cumulative result of the Euclidean distances between all adjacent pixel points on the edge of the connected region. The calculation of the length of the edge of the connected region is well-known; and the larger the value of the edge breakage rate, the stronger the fuzzy feature of the pollution particle region on the surface suppression image or the greater the degree of fuzziness of the pollution particle feature on the surface suppression image.

[0039] After that, calculate the cumulative result of the hole areas of all second target connected regions on the binary image of the surface suppression image, and denote it as the total hole area of the second target connected regions. Calculate the cumulative result of the areas of all second target connected regions on the binary image of the surface suppression image, and denote it as the total area of the suspected pollution particle region. The second target connected region is more likely to be the pollution particle region; then calculate the ratio of the total hole area of the second target connected regions to the total area of the suspected pollution particle region, and use it as the proportion of the hole area of the second target connected regions. And the larger the value of the proportion of the hole area of the second target connected regions, the more holes there are in the second target connected region, and thus the stronger the fuzzy feature of the pollution particle region on the surface suppression image or the greater the degree of fuzziness of the pollution particle feature on the surface suppression image; finally, calculate the mean value of the edge breakage rate and the proportion of the hole area of the second target connected regions, and denote it as the regional fuzzy characterization value, that is, the regional fuzzy characterization value is , where is the broken edge length, is the total edge length, is the total hole area of the second target connected regions, is the total area of the suspected pollution particle region, is the edge breakage rate, is the proportion of the hole area of the second target connected regions; and the larger the regional fuzzy characterization value, the stronger the fuzzy feature of the pollution particle region on the surface suppression image or the greater the degree of fuzziness of the pollution particle feature on the surface suppression image. On the contrary, when the regional fuzzy characterization value is smaller, the degree of fuzziness of the pollution particle feature on the surface suppression image is smaller.

[0040] Therefore, in this embodiment, the global noise characterization value and the regional fuzzy characterization value are obtained through the above process.

[0041] In step S003, based on the global noise characterization value and the regional blur characterization value, a morphological processing strategy corresponding to the surface suppression image is obtained, and the surface suppression image is morphologically processed according to the morphological processing strategy to obtain an image of pollution particles to be detected, and the pollution particles in the image of pollution particles to be detected are detected by using a pollution particle detection model.

[0042] After obtaining the global noise characterization value and the regional blur characterization value, the strategy or method for morphologically processing the surface suppression image is determined based on the values of the global noise characterization value and the regional blur characterization value and the magnitude relationship between the global noise characterization value and the regional blur characterization value. That is, the specific process of obtaining the morphological processing strategy corresponding to the surface suppression image according to the global noise characterization value and the regional blur characterization value is as follows:

[0043] If it is determined that the global noise characterization value is greater than the regional blur characterization value, the global noise characterization value is greater than the preset noise judgment threshold, and the regional blur characterization value is less than the preset regional blur judgment threshold, it indicates that there is more noise in the surface suppression image. Then, a processing method of performing opening operation first and then closing operation should be used to process the surface suppression image to avoid noise interfering with the subsequent closing operation for repair. Therefore, at this time, the morphological processing strategy corresponding to the surface suppression image is to perform opening operation on the surface suppression image first and then perform closing operation.

[0044] If it is determined that the global noise characterization value is less than the regional blur characterization value, the global noise characterization value is less than the preset noise judgment threshold, and the regional blur characterization value is greater than the preset regional blur judgment threshold, it indicates that the blurring of pollution particles on the surface suppression image is relatively serious. Then, a processing method of performing closing operation first and then opening operation should be used to process the surface suppression image to avoid the opening operation further weakening the regional characteristics of pollution particles. Therefore, at this time, the morphological processing strategy corresponding to the surface suppression image is to perform closing operation on the surface suppression image first and then perform opening operation.

[0045] If the global noise characterization value is greater than or equal to the preset noise judgment threshold and the regional blur characterization value is greater than or equal to the preset regional blur judgment threshold, it indicates that both the noise and the blurring of pollution particle characteristics on the surface suppression image are relatively serious. Then, in order to achieve the purpose of blurring repair and denoising of pollution particle characteristics, at this time, a processing method of performing opening operation for denoising first, then closing operation for repair, and finally performing opening operation for smoothing should be used to process the surface suppression image. That is, at this time, the morphological processing strategy corresponding to the surface suppression image is to perform opening operation on the surface suppression image first, then perform closing operation, and then perform opening operation again.

[0046] If it is determined that the global noise characterization value is less than the preset noise judgment threshold and the regional blur characterization value is less than the preset regional blur judgment threshold, it indicates that neither the noise on the surface suppression image nor the degree of blurring of the pollution particle features is serious. Then, in order to avoid the erosion operation from damaging the small particle region features, that is, the pollution particle region features, at this time, the surface suppression image can be directly repaired by a dilation operation. That is, at this time, the dilation operation on the surface suppression image is used as the morphological processing strategy corresponding to the surface suppression image.

[0047] And in specific applications, the implementer needs to set the preset noise judgment threshold and the preset regional blur judgment threshold according to the values of the global noise characterization value and the regional blur characterization value, the impact of noise and the blurring of pollution particle features on the identification of pollution particles, experimental statistics, and other actual situations. For example, in this embodiment, the values of the global noise characterization value and the regional blur characterization value are both from 0 to 1. If the implementer believes that the impact of noise and the blurring of pollution particle features on the identification of pollution particles is the same, then the preset noise judgment threshold and the preset regional blur judgment threshold can both be set to 0.5. If the implementer believes that the blurring of pollution particle features has a greater impact on the identification of pollution particles, then the preset regional blur judgment threshold can be set smaller, such as 0.4.

[0048] After obtaining the morphological processing strategy corresponding to the surface suppression image, the surface suppression image is then morphologically processed according to the morphological processing strategy corresponding to the surface suppression image, and the image obtained after the processing is recorded as the image to be detected for pollution particles. That is, if the morphological processing strategy corresponding to the surface suppression image is to first perform an erosion operation on the surface suppression image and then a dilation operation, then the image to be detected for pollution particles is the image obtained after the surface suppression image is first subjected to an erosion operation and then a dilation operation. After that, a pollution particle detection model is constructed, and the pollution particle detection model is used to detect and identify the pollution particle region on the image to be detected for pollution particles. That is, the image to be detected for pollution particles is input into the pollution particle detection model, and the bounding box and category of the pollution particles on the image to be detected for pollution particles are output.

[0049] Moreover, the pollution particle detection model in this embodiment is a CNN neural network model. The construction of the pollution particle detection model generally includes data preparation, model selection, and training. The input of the pollution particle detection model is an image, and the output is the bounding box and category of the pollution particles. Based on the results output by the model, the bounding box and category label of the pollution particles can be drawn on the image to be detected for pollution particles, and the heat map or distribution density map of the pollution particles can also be obtained. Data preparation generally involves collecting a sample image dataset containing pollution particles. The collected sample image dataset contains images with different illuminations, different angles, and different types of pollution particles, and tools such as LabelImg are used to annotate the positions and categories of the pollution particles. When selecting a model, the implementer generally makes a selection based on the applicable scenarios of different models. For example, models such as Faster R-CNN, YOLO, and SSD are suitable for simultaneous localization and pollution particle classification, and U-Net and Mask R-CNN models are suitable for fine segmentation of pollution particle regions. The training process of the model is well-known and will not be described in detail here. That is, the construction, training, and other processes of the pollution particle detection model in this embodiment are well-known technologies.

[0050] So far, this embodiment has completed the detection and recognition of surface pollution particles on the semiconductor wafer. Moreover, in this embodiment, by performing morphological processing on the surface suppression image, that is, by performing fuzzy feature repair and denoising on the surface suppression image, it is possible to improve the accuracy and integrity of pollution particle recognition while reducing the influence of interference features such as circuit patterns or lithography structures on pollution particle recognition, and at the same time, it is possible to minimize the influence of the existence of pollution particle feature blurring on the accuracy and integrity of pollution particle recognition, thereby improving the accuracy and integrity of detecting and recognizing pollution particles on the surface of the semiconductor wafer.

[0051] In summary, in this embodiment, first, a surface suppression image of a semiconductor wafer to be detected is obtained; then, a preset window is used to divide the surface suppression image to obtain all local windows on the surface suppression image, the first target connected domain and the second target connected domain on the binary image of the surface suppression image are obtained, and a global noise characterization value is obtained according to the pixel value variance in the local window and the area ratio of the first target connected domain on the binary image; then, the broken connected domains in all the second target connected domains are obtained, and a regional blur characterization value is obtained according to the total edge length of all the broken connected domains, the total edge length of all the second target connected domains, and the hole area of the second target connected domain; finally, according to the global noise characterization value and the regional blur characterization value, a morphological processing strategy corresponding to the surface suppression image is obtained, and the surface suppression image is morphologically processed according to the morphological processing strategy to obtain an image of contamination particles to be detected, and a contamination particle detection model is used to detect and identify the contamination particles in the image of contamination particles to be detected. Moreover, in this embodiment, morphological processing is performed on the surface suppression image, that is, by repairing the blur features of the surface suppression image, the influence of the existence of the contamination particle feature blur phenomenon on the accuracy and integrity of contamination particle recognition can be reduced as much as possible, thereby improving the accuracy and integrity of detecting and identifying the contamination particles on the surface of the semiconductor wafer.

[0052] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. An image detection method for particle contamination on the surface of a semiconductor wafer, characterized in that, The method includes the following steps: Obtain a surface suppression image of a semiconductor wafer to be detected; Use a preset window to divide the surface suppression image to obtain all local windows on the surface suppression image, obtain a first target connected component and a second target connected component on the binary image of the surface suppression image, and obtain a global noise characterization value according to the pixel value variance in the local window and the area ratio of the first target connected component on the binary image; Obtain the broken connected components in all the second target connected components, and obtain a regional blur characterization value according to the total edge length of all the broken connected components, the total edge length of all the second target connected components, and the hole area of the second target connected components; According to the global noise characterization value and the regional blur characterization value, obtain a morphological processing strategy corresponding to the surface suppression image, perform morphological processing on the surface suppression image according to the morphological processing strategy to obtain an image of pollution particles to be detected, and use a pollution particle detection model to detect pollution particles in the image of pollution particles to be detected.

2. The image detection method for particle contamination on the surface of a semiconductor wafer according to claim 1, wherein The method for obtaining the first target connected component and the second target connected component includes: Perform connected component analysis on the binary image, and denote the obtained connected components as the connected components to be analyzed on the binary image; Judge whether the area of the connected component to be analyzed is less than a preset connected component area threshold. If so, denote the corresponding connected component to be analyzed as the first target connected component, otherwise, denote the corresponding connected component to be analyzed as the second target connected component.

3. The image detection method for particle contamination on the surface of a semiconductor wafer according to claim 1, wherein, The method for obtaining the global noise characterization value includes: Normalize the variance of the pixel values of all pixel points in the local window, and denote it as the normalized variance corresponding to the local window; Obtain a first noise feature value according to the normalized variance corresponding to the local window; Obtain a second noise feature value according to the ratio of the sum of the areas of all the first target connected components on the binary image to the total number of pixel points in the binary image; Calculate the weighted sum result of the first noise feature value and the second noise feature value, and denote it as the global noise characterization value.

4. The image detection method for particle contamination on the surface of a semiconductor wafer according to claim 3, wherein, The method for obtaining the first noise feature value includes: Obtain the number of local windows with a normalized variance greater than a preset variance threshold, and denote it as the number characterization value, calculate the ratio of the number characterization value to the total number of local windows on the surface suppression image, and denote it as the feature ratio, and normalize the feature ratio, and denote it as the first noise feature value.

5. The image detection method for particle contamination on the surface of a semiconductor wafer according to claim 3, wherein, The method for obtaining the second noise feature value includes: Denote the ratio of the sum of the areas of all the first target connected components on the binary image to the total number of pixel points in the binary image as the area ratio of the first target connected component. If the area ratio of the first target connected component is less than a preset ratio threshold, use the constant 0 as the second noise characterization value. If the area ratio of the target connected component is greater than the preset ratio threshold, use the constant 1 as the second noise characterization value.

6. The image detection method for particle contamination on the surface of a semiconductor wafer according to claim 1, wherein, The method for obtaining the broken connected component includes: For any second target connected domain, use an edge detection algorithm to extract the edge of the second target connected domain to obtain the edge of the second target connected domain, and determine whether the edge of the second target connected domain is a non-closed edge. If so, mark the second target connected domain as a broken connected domain.

7. The image detection method for particle contamination on the surface of a semiconductor wafer according to claim 1, wherein The method for obtaining the hole area of the second target connected domain includes: For any second target connected domain, fill the second target connected domain and mark the filled second target connected domain as the filled connected domain corresponding to the second target connected domain. Denote the result of subtracting the area of the filled connected domain corresponding to the second target connected domain from the area of the second target connected domain as the hole area of the second target connected domain.

8. The image detection method for particle contamination on the surface of a semiconductor wafer according to claim 1, characterized in that, The method for obtaining the regional fuzzy characterization value includes: Calculate the sum of the edge lengths of all broken connected domains on the binary image and denote it as the broken edge length. Calculate the sum of the edge lengths of all second target connected domains on the binary image and denote it as the total edge length. Obtain the edge breakage rate based on the broken edge length and the total edge length; Calculate the cumulative result of the hole areas of all second target connected domains on the binary image and denote it as the total hole area. Calculate the cumulative result of the areas of all second target connected domains on the binary image and denote it as the total area of the suspected contaminated particle region; Use the ratio of the total hole area to the total area of the suspected contaminated particle region as the proportion of the hole area of the second target connected domain; Calculate the mean of the edge breakage rate and the proportion of the hole area of the second target connected domain and denote it as the regional fuzzy characterization value.

9. The image detection method for particle contamination on the surface of a semiconductor wafer according to claim 8, wherein, The edge breakage rate is the ratio of the broken edge length to the total edge length.

10. The image detection method for particle contamination on the surface of a semiconductor wafer according to claim 1, characterized in that, The method for obtaining the morphological processing strategy corresponding to the surface suppression image based on the global noise characterization value and the regional fuzzy characterization value includes: If the global noise characterization value is greater than the regional fuzzy characterization value, the global noise characterization value is greater than the preset noise judgment threshold, and the regional fuzzy characterization value is less than the preset regional fuzzy judgment threshold, then use performing an opening operation on the surface suppression image first and then a closing operation as the morphological processing strategy corresponding to the surface suppression image; If the global noise characterization value is less than the regional fuzzy characterization value, the global noise characterization value is less than the preset noise judgment threshold, and the regional fuzzy characterization value is greater than the preset regional fuzzy judgment threshold, then use performing a closing operation on the surface suppression image first and then an opening operation as the morphological processing strategy corresponding to the surface suppression image; If the global noise characterization value is greater than or equal to the preset noise judgment threshold and the regional fuzzy characterization value is greater than or equal to the preset regional fuzzy judgment threshold, then use performing an opening operation on the surface suppression image first, then a closing operation, and then an opening operation again as the morphological processing strategy corresponding to the surface suppression image; If the global noise characterization value is less than a preset noise judgment threshold and the regional blur characterization value is less than a preset regional blur judgment threshold, then performing a closing operation on the surface suppression image is used as the morphological processing strategy corresponding to the surface suppression image.

Images