Sample matching method, sample matching system and storage medium

Abstract

Provided in the present invention are a sample matching method, a sample matching system and a storage medium. The sample matching method comprises the following steps: acquiring a first defect point set of a sample under test, and a feature point set of at least one standard sample, wherein the first defect point set and the feature point set each comprise a plurality of elements, and each element records a coordinate position of a corresponding point in a corresponding sample; performing downsampling processing on the first defect point set, so as to construct a second defect point set, and searching the feature point set for a first element corresponding to each second element in the second defect point set, so as to determine a plurality of candidate point pairs; screening the candidate point pairs on the basis of a preset first distance threshold value, so as to determine a plurality of registration point pairs; and on the basis of the plurality of registration point pairs, determining a matching result of the sample under test and the standard sample. The present invention can be used for improving the matching rate and matching speed of sample matching, so as to further improve the matching efficiency of a defect detection device.

Description

Sample matching method, sample matching system and storage medium TECHNICAL FIELD

[0001] The present application relates to a semiconductor defect detection technology, in particular to a sample matching method, a sample matching system and a computer readable storage medium. BACKGROUND

[0002] In a defect detection device, due to the large amount of image data, especially in the process of high-resolution image acquisition, the registration and processing of the image will face some challenges. In this case, the image is usually first matched with samples, that is, by extracting the significant feature points such as corner points, edge points and interest points in the image, a robust correspondence between images is established to achieve accurate image registration. Since the registration is performed by extracting feature points, the sample matching problem of the semiconductor standard sheet can be regarded as a point set registration problem. The point set registration method is divided into rigid registration and non-rigid registration. Rigid registration produces a rigid transformation, which does not change the distance between any two points, that is, the shape and size of the whole do not change, and generally such transformation only includes translation and rotation. Non-rigid registration produces a non-rigid transformation, which can change the distance between two points, that is, the shape and size of the whole can change, including affine transformation, such as scaling and shearing, etc., and can also involve other nonlinear transformations.

[0003] The currently widely used is the Iterative Closest Point (ICP) algorithm, which can gradually optimize the matching points through iteration to approach the optimal solution, so as to realize high-precision point cloud registration. Its principle is simple and easy to understand, and easy to operate. However, the ICP algorithm is sensitive to the initial position of the point set, and may fall into a local optimal solution, so it is only suitable for rigid transformation scenes, but not for non-rigid transformation scenes. The emergence of the Coherent Point Drift (CPD) algorithm to a certain extent makes up for the shortcomings of the ICP algorithm. It is more widely applicable, and can handle rigid and non-rigid transformation scenes. It is not sensitive to the initial position of the point set, has strong robustness, and can achieve a globally optimal registration result. However, the CPD algorithm has high computational complexity, and the defect point set of the standard sheet collected in the actual defect detection has many noise points. Direct use of the CPD algorithm for registration is time-consuming and has low matching rate.

[0004] In order to overcome the above-mentioned defects existing in the prior art, there is an urgent need in the field for an improved sample matching method for improving the matching rate and matching speed of sample matching, so as to further improve the matching efficiency of the defect detection device.

[0005] SUMMARY

[0006] The following presents a simplified summary of one or more aspects in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects, and is intended to neither identify key or critical elements of all aspects nor delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.

[0007] In order to overcome the above-mentioned defects existing in the prior art, the present application provides a sample matching method, a sample matching system and a computer readable storage medium, which can improve the matching rate and speed of sample matching by optimizing the defect point set and then using the CPD algorithm for matching, so as to further improve the matching efficiency of the defect detection device.

[0008] Specifically, according to the sample matching method provided by the first aspect of the present application, the method comprises the following steps: obtaining a first defect point set of a to-be-tested sample and a feature point set of at least one standard sample. The first defect point set and the feature point set each comprise a plurality of elements, and each element records the coordinate position of a corresponding point in a corresponding sample; performing down-sampling processing on the first defect point set to construct a second defect point set, and searching for a first element corresponding to each second element in the second defect point set in the feature point set to determine a plurality of candidate point pairs; screening each candidate point pair according to a preset first distance threshold to determine a plurality of registration point pairs; and determining a matching result of the to-be-tested sample and the standard sample according to the plurality of registration point pairs.

[0009] Further, in some embodiments of the present application, the step of performing down-sampling processing on the first defect point set to construct a second defect point set comprises: determining a grid interval for each coordinate axis according to the value range of each third element in each coordinate axis in the first defect point set and a preset resolution; and drawing a grid according to the grid interval and extracting one element in each grid unit to construct the second defect point set.

[0010] Further, in some embodiments of the present application, the step of searching for a first element corresponding to each second element in the second defect point set in the feature point set to determine a plurality of candidate point pairs comprises: mapping each first element in the original first feature point set and each second element in the second defect point set to a k-dimensional space respectively, and determining the nearest first element of each second element in the k-dimensional space to construct a second feature point set; and determining a plurality of first candidate point pairs according to each fourth element in the second feature point set and the second element corresponding thereto.

[0011] Further, in some embodiments of the present application, the step of screening each of the candidate point pairs according to the preset first distance threshold to determine a plurality of registration point pairs comprises: calculating the distance between the second element and the fourth element in each of the first candidate point pairs, respectively; and eliminating the first candidate point pairs with a distance greater than or equal to the first distance threshold, and retaining a plurality of first candidate point pairs with a distance less than the first distance threshold as the first registration point pairs.

[0012] Further, in some embodiments of the present application, the step of determining the matching result of the test sample and the standard sample according to the plurality of registration point pairs comprises: determining a rotation matrix and a translation matrix representing the relative positional relationship between the second feature point set and the second defect point set according to the plurality of first registration point pairs; processing the first feature point set according to the rotation matrix and the translation matrix to construct a third feature point set; searching for a fifth element corresponding to each of the second elements in the second defect point set in the third feature point set to determine a plurality of second candidate point pairs; screening each of the second candidate point pairs according to a preset second distance threshold to determine a plurality of second registration point pairs, wherein the second distance threshold is less than the first distance threshold; and calculating the ratio of the number of the second registration point pairs to the number of the first elements in the first feature point set to determine the matching result of the test sample and the standard sample.

[0013] Further, in some embodiments of the present application, the step of determining a rotation matrix and a translation matrix representing the relative positional relationship between the second feature point set and the second defect point set according to the plurality of first registration point pairs comprises: splitting the set Z1 = {(x1', y1), (x2', y2),..., (x k ′, y k )} of the first registration point pairs into a point set M = {x1', x2',..., x k ′} about the second elements and a point set S = {y1, y2,..., y k} about the fourth elements, and representing the distribution of the point set M as a mixture Gaussian model. The negative log-likelihood function of the probability density estimation problem representing the point set registration result reaches a minimum when the point set M and the point set S are completely registered; each point x i ′ in the point set M is taken as the center of each component in the mixture Gaussian model, and the probability of each component is assumed to be equal to obtain a corresponding first distribution model:

[0014] wherein,

[0015] P(i) is a mixing ratio coefficient, representing a prior probability of the i-th Gaussian distribution. p(S|i) represents a probability density function of the i-th Gaussian distribution. σ 2 is a variance of the i-th Gaussian distribution. D represents a dimension of a point set, and takes a value of 2; and 2 minimizing the negative log-likelihood function to determine model parameters θ and σ 2 of the mixture Gaussian model. The negative log-likelihood function is represented as:

[0016] θ represents a set of all parameters in the spatial transformation model T, including the rotation matrix R and the translation matrix t.

[0017] Further, in some embodiments of the present application, before minimizing the negative log-likelihood function, the sample matching method further comprises the following step: adding a uniform distribution function to the first distribution model to optimize the first distribution model into a second distribution model containing noise effects:

[0018] wherein w is a weight of the uniform distribution function.

[0019] Further, in some embodiments of the present application, the step of minimizing the negative log-likelihood function to determine model parameters θ and σ 2 of the mixture Gaussian model comprises: using an expectation-maximization algorithm to first calculate a posterior probability distribution P old (i|y j ) by using old distribution model parameters θ old and σ old , and then minimizing the negative log-likelihood function according to the posterior probability distribution P old (i|y j ) to iteratively determine new model parameters θ and σ 2 , wherein the posterior probability distribution P old (i|y j ) is represented as:

[0020] The negative log-likelihood function is further represented as:

[0021] Further, in some embodiments of the present application, the step of searching the third feature point set for a fifth element corresponding to each second element in the second defect point set to determine a plurality of second candidate point pairs comprises: mapping each fifth element in the third feature point set and each second element in the second defect point set to a k-dimensional space, respectively, and determining a nearest fifth element to each second element in the k-dimensional space to construct a fourth feature point set; and determining the plurality of second candidate point pairs according to each sixth element in the fourth feature point set and its corresponding second element.

[0022] Further, in some embodiments of the present application, the second distance threshold is half of the first distance threshold, and the step of screening each second candidate point pair according to a preset second distance threshold to determine a plurality of second registration point pairs comprises: calculating a distance between a second element and a sixth element in each second candidate point pair, respectively; and eliminating a second candidate point pair whose distance is greater than or equal to the second distance threshold, and retaining a plurality of second candidate point pairs whose distance is less than the second distance threshold as the second registration point pairs.

[0023] In addition, the sample matching system provided by the second aspect of the present application comprises a memory and a processor. The memory has computer instructions stored thereon. The processor is connected to the memory and is configured to execute the computer instructions stored on the memory to implement the sample matching method provided by the first aspect of the present application.

[0024] In addition, the computer readable storage medium provided by the third aspect of the present application has computer instructions stored thereon. The computer instructions are executed by a processor to implement the sample matching method provided by the first aspect of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0025] The above features and advantages of the present application can be better understood by reading the following detailed description of embodiments of the present application in conjunction with the drawings, in which: the components are not necessarily drawn to scale, and components of similar or identical function or features can have the same or similar reference label.

[0026] FIG. 1 shows a flowchart of a sample matching method according to some embodiments of the present application.

[0027] FIG. 2 shows a schematic diagram of a feature point set of a labeled sample according to some embodiments of the present application.

[0028] FIG. 3 shows a schematic diagram of a first defect point set of a sample to be tested according to some embodiments of the present application.

[0029] FIG. 4 shows a schematic diagram of a second set of defect points after down-sampling processing, according to some embodiments of the present application.

[0030] FIG. 5 shows a schematic diagram of a first pair of registration points after screening, according to some embodiments of the present application.

[0031] FIG. 6 shows a schematic diagram of a matching result, according to some embodiments of the present application. DETAILED DESCRIPTION

[0032] Those skilled in the art will readily understand that the application is well adapted to carry out the objects and obtain the ends and advantages mentioned, as well as those inherent therein. While the application has been depicted, described and is defined by reference to preferred embodiments of the application, such references do not imply a limitation on the application, and no such limitation is to be inferred. The application is capable of considerable modification, alteration, and equivalents in form and function, without departing from the scope of the application as described in the claims.

[0033] In the description of the present application, it is necessary to explain that, unless explicitly defined and limited, the terms "mounting", "connection", "linking" should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be direct connection, or indirect connection through intermediate medium, or internal connection of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0034] In addition, in the following description, "up", "down", "left", "right", "top", "bottom", "horizontal", "vertical" should be understood as the orientation shown in the paragraph and the related drawings. The relative terms are only used for the convenience of description, and they do not mean that the devices described should be manufactured or operated in a particular orientation, so they should not be understood as a limitation on the application.

[0035] It is understood that although the terms "first", "second", "third" and the like can be used herein to describe various components, regions, layers and / or sections, these components, regions, layers and / or sections should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers and / or sections. Therefore, the first component, region, layer and / or section discussed below can be referred to as the second component, region, layer and / or section without departing from some embodiments of the present application.

[0036] As mentioned above, the ICP algorithm is sensitive to the initial position of the point set and may fall into a local optimum. Therefore, it is only suitable for scenarios with rigid transformations, but not for those with non-rigid transformations. The CPD algorithm has high computational complexity, and the defect point sets collected from standard films during actual defect detection often contain a lot of noise. Directly using the CPD algorithm for registration is time-consuming and has a low matching rate.

[0037] In order to overcome the above-mentioned defects of the prior art, the present invention provides a sample matching method, a sample matching system and a computer-readable storage medium, which can improve the matching rate and matching speed of sample matching by first optimizing the defect point set and then matching it using the CPD algorithm, so as to further improve the matching efficiency of the defect detection equipment.

[0038] In some non-limiting embodiments, the sample matching method provided in the first aspect of the present invention can be implemented based on the sample matching system provided in the second aspect of the present invention. Specifically, the sample matching system provided in the second aspect of the present invention includes a memory and a processor. Here, the memory includes, but is not limited to, the computer-readable storage medium provided in the third aspect of the present invention, having computer instructions stored thereon. The processor is connected to the memory and configured to execute the computer instructions stored in the memory to implement the sample matching method provided in the first aspect of the present invention.

[0039] For details, please refer to Figures 1 to 6. Figure 1 shows a flow chart of a sample matching method provided according to some embodiments of the present invention. Figure 2 shows a schematic diagram of a feature point set of annotated samples provided according to some embodiments of the present invention. Figure 3 shows a schematic diagram of a first defect point set of a sample to be tested provided according to some embodiments of the present invention. Figure 4 shows a schematic diagram of a second defect point set after downsampling processing provided according to some embodiments of the present invention. Figure 5 shows a schematic diagram of a first registration point pair after screening provided according to some embodiments of the present invention. Figure 6 shows a schematic diagram of a matching result provided according to some embodiments of the present invention.

[0040] As shown in FIG1 to FIG3, during the sample matching process, the processor may first obtain the first defect point set Y={y1, y2, ..., y n}, and at least one feature point set X={x1,x2,...,x m}. Here, each element x m and y n The coordinate position (x, y) of a corresponding point in the corresponding sample is recorded. Since the defects of the standard sample are set, the amount of data in its feature point set X is fixed, while the amount of data in the first defect point set Y of the sample to be tested is not fixed. Therefore, the amount of data in X and Y is not necessarily equal.

[0041] Please refer to FIG. 1 and FIG. 4, after obtaining the first defect point set Y and the feature point set X of at least one standard sample, the processor can down-sample the first defect point set Y to construct the second defect point set shown in FIG. 4.

[0042] Specifically, the processor can first determine the grid interval of each coordinate axis according to the value range of each third element in the first defect point set Y on each coordinate axis and the preset resolution, as follows:

[0043] X_ave=(X_max-X_min) / M

[0044] Y_ave=(Y_max-Y_min) / N

[0045] Wherein, X_ave and Y_ave represent the grid interval of x coordinate and y coordinate, X_max and X_min represent the maximum and minimum value of x coordinate in the first defect point set Y, Y_max and Y_min represent the maximum and minimum value of y coordinate in the first defect point set Y, M and N represent the preset resolution of x coordinate and y coordinate, respectively. The setting of the value will affect the result. Therefore, during the down-sampling process, the value of M and N needs to be debugged. If the matching result can reach more than eighty percent, it is determined that the value setting is appropriate.

[0046] Then, the processor can draw a grid of N rows and M columns according to the above grid interval x_ave and Y_ave, and extract an element in each grid unit to construct the second defect point set Y1={y1,y2,...,y s}(s≤n). In this way, by down-sampling the first defect point set Y, the calculation amount of the processor is reduced, and the influence of noise points on the result is reduced.

[0047] Then, as shown in FIG. 1, the processor can search for the first element corresponding to each second element in the second defect point set Y1 in the feature point set X to determine a plurality of candidate point pairs.

[0048] Specifically, the processor can first use the K-Nearest Neighbors (KNN) algorithm to map each first element in the original first feature point set X and each second element in the second defect point set Y1 to a k-dimensional space, respectively, and determine the nearest first element of each second element in the k-dimensional space to construct the second feature point set X1={x′1,x′2,...,x′ s}。Afterwards, the processor can determine a plurality of first candidate point pairs Z = {(x1', y1), (x2', y2),..., (x s ′,y s )} according to each fourth element in the second feature point set X1 and its corresponding second element. Here, the KNN algorithm uses a KD tree structure to efficiently search the nearest neighbor first element in a K-dimensional space to speed up the search process of the nearest neighbor point. In addition, since the nearest neighbor point found by the algorithm can be a noise point rather than the corresponding first element for a certain second element, the first candidate point pair Z obtained above is only theoretically optimal.

[0049] Please refer to Figures 1 and 5, after determining the point pair set Z composed of a plurality of candidate point pairs, the processor can screen each candidate point pair in the point pair set Z according to a preset first distance threshold d1 to determine a plurality of registration point pairs. Here, the value of the first distance threshold d1 is set according to the actual situation.

[0050] Specifically, the processor can calculate the distance between the second element and the fourth element in each first candidate point pair Z:

[0051] where x a , x b , y a , y b represent the x and y coordinates of the first candidate point pair a and b, respectively.

[0052] Further, in some embodiments, the first distance threshold d1 can be preferably set to 20um. The processor can eliminate the first candidate point pair whose distance is greater than or equal to the first distance threshold d1, and retain a plurality of first candidate point pairs whose distance is less than the first distance threshold d1 as the first registration point pair, and determine the corresponding first registration point pair set Z1 = {(x1', y1), (x2', y2),..., (x k ′,y k ))(k≤s).

[0053] Afterwards, as shown in Figure 1, the processor can determine the matching result of the test sample and the standard sample according to the plurality of registration point pairs in the first registration point pair set Z1.

[0054] In some embodiments, the processor can first determine the rotation matrix R and the translation matrix t representing the relative positional relationship between the second feature point set and the second defect point set according to the plurality of first registration point pairs Z1, and determine the matching result of the test sample and each standard sample based on the rotation matrix R and the translation matrix t.

[0055] Specifically, the processor can utilize the CPD algorithm to split the above-mentioned first set of point pairs Z1 = {(x1', y1), (x2', y2),..., (x k k ), (x k ', y k )} into a point set M = {x1', x2',..., x i '} about the second element and a point set S = {y1, y2,..., y 2} about the fourth element, and express the distribution of the point set M as a mixture Gaussian model. Here, the processor can utilize the CPD algorithm to convert the matching problem of the point set M and the point set S into an estimation problem of the probability density, i.e., the negative log-likelihood function of the probability density estimation problem representing the registration result of the point set reaches the minimum when the point set M and the point set S are registered.

[0056] Subsequently, the processor can take each point x 2 ' in the point set M as the center of each component in the mixture Gaussian model, and assume that the probability of each component is equal to obtain a first distribution model corresponding to the linear superposition of k Gaussian distribution models:

[0057] wherein,

[0058] wherein, P(i) is a mixture proportion coefficient, representing the prior probability of the i-th Gaussian distribution. p(S|i) represents the probability density function of the i-th Gaussian distribution, σ 2 is the variance of the i-th Gaussian distribution, and D represents the dimension of the point set, taking the value of 2.

[0059] In addition, in some preferred embodiments, the processor can also add a uniform distribution function to the above-mentioned first distribution model to optimize the first distribution model into a second distribution model containing the influence of noise:

[0060] wherein, w is the weight of the uniform distribution function.

[0061] Further, the processor can minimize the above-mentioned negative log-likelihood function to determine the model parameters θ and σ 2 of the mixture Gaussian model. Here, the negative log-likelihood function is expressed as:

[0062] wherein, θ represents the set of all parameters in the spatial transformation model T, including the rotation matrix R and the translation matrix t.

[0063] Further, the processor can employ the Expectation Maximum (EM) iterative optimization algorithm to optimize the negative log-likelihood function L(θ, σ 2 .). Here, the maximum expectation algorithm can first pass the old distribution model parameters θ old and σ old Calculate the posterior probability distribution P old (i|y j ), and then according to the posterior probability distribution P ol3 (i|y j ) minimizes the negative log-likelihood function to iteratively determine the new model parameters θ and σ 2 .

[0064] Here, the posterior probability distribution P old (i|y j ) is represented as:

[0065] The negative log-likelihood function is further expressed as:

[0066] Thereafter, the processor may determine the rotation matrix R and the translation matrix t according to the model parameters θ determined above, and then process the first feature point set X according to the rotation matrix R and the translation matrix t to construct a third feature point set X′:

[0067] X′=X·R+t

[0068] Thereafter, the processor may search the third feature point set X′ for the fifth elements corresponding to the second elements in the second defect point set Y1 to determine a plurality of second candidate point pairs, and then filter the second candidate point pairs according to a preset second distance threshold to determine a plurality of second registration point pairs.

[0069] Similarly, the processor can use the KNN algorithm to map each fifth element in the third feature point set and each second element in the second defect point set Y1 to k-dimensional space and determine the fifth element closest to each second element in k-dimensional space to construct a fourth feature point set. The processor can then determine multiple second candidate point pairs based on each sixth element in the fourth feature point set and its corresponding second element. Similarly, since the nearest neighbor point found by this algorithm for a particular second element may be a noise point rather than the corresponding fifth element, the second candidate point pairs obtained above are only theoretically optimal.

[0070] Then, the processor can screen the second candidate point pairs according to a preset second distance threshold d2 to determine multiple second registration point pairs. Here, the second distance threshold d2 is smaller than the first distance threshold d1 so that technicians can quickly obtain results and reduce the time of debugging parameters.

[0071] Further, in some embodiments, the second distance threshold d2 is preferably 0.5 times the first distance threshold d1. In the process of screening each second candidate point pair, the processor can calculate the distance between the second element and the sixth element in each second candidate point pair, respectively:

[0072] wherein x a , x b , y a , y b represent the x coordinate and the y coordinate of the second candidate point pair of a and b, respectively.

[0073] Then, the processor can eliminate the second candidate point pair with a distance greater than or equal to the second distance threshold d2, and retain the second candidate point pairs with a distance less than the second distance threshold d2 as the second registration point pairs.

[0074] Further, as shown in FIG. 6, the processor can calculate the ratio of the number of the second registration point pairs to the number of the first elements in the first feature point set to determine the matching result of the test sample and the standard sample.

[0075] For example, if the ratio of the number of the second registration point pairs to the number of the first elements in the first feature point set is greater than or equal to a preset threshold (e.g., 95%), the processor can determine that the test sample matches the standard sample. Conversely, if the ratio of the number of the second registration point pairs to the number of the first elements in the first feature point set is less than the preset threshold, the processor can determine that the test sample does not match the standard sample.

[0076] In summary, the sample matching method, the sample matching system, and the computer readable storage medium provided by the present application can improve the matching rate and the matching speed of sample matching by optimizing the defect point set and then matching by using the CPD algorithm, so as to further improve the matching efficiency of the defect detection device.

[0077] Although the above-described methods are illustrated and described as a series of acts, it will be appreciated and understood by those skilled in the art that the methods are not limited by the order of the acts, as some acts can occur in different orders and / or concurrently with other acts according to one or more embodiments.

[0078] Those skilled in the art will understand that information, signals, and data can be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that can be referenced throughout the above description can be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0079] Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans can implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

[0080] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples described herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A sample matching method, characterized by, The method comprises the following steps: obtaining a first defect point set of a sample to be tested and a feature point set of at least one standard sample, wherein the first defect point set and the feature point set each comprise a plurality of elements, and each element records a corresponding point in a corresponding sample; performing down-sampling processing on the first defect point set to construct a second defect point set, and searching for a first element corresponding to each second element in the second defect point set in the feature point set to determine a plurality of candidate point pairs; screening each candidate point pair according to a preset first distance threshold to determine a plurality of registration point pairs; and determining a matching result of the sample to be tested and the standard sample according to the plurality of registration point pairs.

2. The sample matching method of claim 1, wherein, The step of performing down-sampling processing on the first defect point set to construct a second defect point set comprises: determining a grid interval for each coordinate axis according to a value range of each third element in the first defect point set on each coordinate axis and a preset resolution; and drawing a grid according to the grid interval, and extracting one element in each grid cell to construct the second defect point set.

3. The sample matching method of claim 1, wherein, The step of searching for a first element corresponding to each second element in the second defect point set in the feature point set to determine a plurality of candidate point pairs comprises: mapping each first element in an original first feature point set and each second element in the second defect point set to a k-dimensional space respectively, and determining a nearest first element of each second element in the k-dimensional space to construct a second feature point set; and determining a plurality of first candidate point pairs according to each fourth element in the second feature point set and a corresponding second element.

4. The sample matching method of claim 3, wherein, The step of screening each candidate point pair according to a preset first distance threshold to determine a plurality of registration point pairs comprises: calculating a distance between a second element and a fourth element in each first candidate point pair respectively; and eliminating a first candidate point pair with a distance greater than or equal to the first distance threshold, and retaining a plurality of first candidate point pairs with a distance less than the first distance threshold as first registration point pairs.

5. The sample matching method of claim 4, wherein, The step of determining a matching result of the sample to be tested and the standard sample according to the plurality of registration point pairs comprises: determining a rotation matrix and a translation matrix representing a relative positional relationship between the second feature point set and the second defect point set according to the plurality of first registration point pairs; processing the first feature point set according to the rotation matrix and the translation matrix to construct a third feature point set; searching for a fifth element corresponding to each second element in the second defect point set in the third feature point set to determine a plurality of second candidate point pairs; screening each second candidate point pair according to a preset second distance threshold to determine a plurality of second registration point pairs, wherein the second distance threshold is less than the first distance threshold; and calculating a ratio of a number of the second registration point pairs to a number of the first elements in the first feature point set to determine the matching result of the sample to be tested and the standard sample.

6. The sample matching method of claim 5, wherein, The step of determining a rotation matrix and a translation matrix representing the relative positional relationship between the second feature point set and the second defect point set according to the plurality of first candidate point pairs comprises: a set of pairs of first registration points Z1 = {(x1', y1), (x2', y2),..., (x k ', y k )} is split into a point set M = {x1', x2',..., x k '} with respect to the second element and a point set S = {y1, y2,..., y k} with respect to the fourth element and the distribution of the point set M is represented as a mixture of Gaussians, wherein a negative log-likelihood function characterizing a probability density estimation problem of the point set registration results reaches a minimum when the point set M is registered with the point set S; Each point x in the set M is taken as the center of each component in the mixture Gaussian model, and the probability of each point x in the set M is calculated as follows: i ′ Each point x in the set M is taken as the center of each component in the mixture Gaussian model, and the probability of each point x in Let each component probability be equal to obtain the corresponding first distribution model: wherein P(i) is a mixing ratio coefficient, representing a prior probability of the ith Gaussian distribution, p(S|i) represents a probability density function of the ith Gaussian distribution, σ 2 is a variance of the ith Gaussian distribution, D represents a dimension of the point set, and takes a value of 2; and minimizing the negative log-likelihood function to determine model parameters θ and σ of the mixture Gaussian model 2 wherein the negative log-likelihood function is represented as: θ represents all parameter sets in the spatial transformation model T, including the rotation matrix R and the translation matrix t.

7. The sample matching method of claim 6, wherein, Before minimizing the negative log-likelihood function, the sample matching method further comprises the following steps: adding a uniform distribution function to the first distribution model to optimize the first distribution model to a second distribution model that includes the effect of noise: wherein w is the weight of the uniform distribution function.

8. The sample matching method of claim 6, wherein, minimizing the negative log-likelihood function to determine model parameters θ and σ of the mixture Gaussian model 2 comprises: Using the expectation-maximization algorithm, the posterior probability distribution is first computed using the old model parameters θ old and σ old ​ P old (i|y j ), and minimizing the negative log-likelihood function based on the posterior probability distribution P old (i|y j ) to iteratively determine new model parameters θ and σ 2 where the posterior probability distribution P old (i|y j ) is represented as: The negative log-likelihood function is further expressed as:

9. The sample matching method of claim 5, wherein, The step of searching for a fifth element corresponding to each second element in the second defect point set in the third feature point set to determine a plurality of second candidate point pairs comprises: mapping each fifth element in the third feature point set and each second element in the second defect point set to a k-dimensional space respectively, and determining the nearest fifth element of each second element in the k-dimensional space to construct a fourth feature point set; and determining the plurality of second candidate point pairs according to each sixth element in the fourth feature point set and its corresponding second element.

10. The sample matching method of claim 9, wherein, The second distance threshold is half of the first distance threshold, and the step of screening each second candidate point pair according to a preset second distance threshold to determine a plurality of second candidate point pairs comprises: calculating the distance between the second element and the sixth element in each second candidate point pair respectively; and eliminating the second candidate point pair whose distance is greater than or equal to the second distance threshold, and retaining a plurality of second candidate point pairs whose distance is less than the second distance threshold as the second candidate point pair.

11. A sample matching system, characterized by, comprises: a memory having computer instructions stored thereon; and a processor connected to the memory and configured to execute the computer instructions stored on the memory to implement the sample matching method according to any one of claims 1-10. The computer instructions are executed by the processor to implement the sample matching method according to any one of claims 1-10.

12. A computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are executed by the processor to implement the sample matching method according to any one of claims 1-10.

Images