Detection formula setting and optimization method and device, electronic equipment and storage medium

Abstract

The application provides a detection formula setting and optimization method and device, electronic equipment and a storage medium. The method comprises the following steps: labeling a first data sample to obtain a second data sample; wherein the first data sample comprises a plurality of detection result data; the second data sample comprises the detection result data and a label corresponding to each piece of data; and data feature distribution information of a detection object is obtained according to the second data sample; a preset outlier statistical analysis strategy is used to perform outlier statistical analysis on the data feature distribution information, to obtain defect distribution boundary information and determine a detection formula; and finally, the values of the detection parameters of the detection formula are determined or optimized through reverse deduction according to the defect distribution boundary information and the preset outlier statistical analysis strategy. The application considers the coupling relationship between parameters, can avoid repeated adjustment of parameters, can infer a complete set of detection parameters at the same time, realizes rapid modeling of the detection formula, and can save manpower and time costs.

Description

Technical Field

[0001] This invention relates to the field of semiconductor technology, and in particular to a method, apparatus, electronic device, and storage medium for detecting formulation setting and optimization. Background Technology

[0002] In the semiconductor wafer manufacturing process, wafer warpage and surface morphology are key parameters affecting process stability and product yield, significantly impacting wafer yield. For example, after different processes such as etching or thin film deposition, wafers may warp to varying degrees or develop uneven surfaces. Furthermore, robotic arms may scratch wafers during semiconductor integrated circuit manufacturing. Therefore, wafer defects are a primary concern for all chip manufacturers in yield inspection. Once wafer defects exist, they are difficult to correct through subsequent processes. Therefore, the ability to quickly and accurately detect wafer surface defects and prevent defective products from flowing into the next process, thus avoiding wasted production resources, is crucial.

[0003] In existing technologies, wafer defect detection processes typically employ a forward process for parameter tuning. However, due to the diversity of on-site processes, a large amount of information needs to be generated each time. Coupled with a lack of prior knowledge, the detection parameters in the process are usually adjusted one by one. Since the coupling relationships between parameters cannot be taken into account, repeated adjustments to a single parameter may lead to deviations in the tuning results. To achieve better detection results, the detection formula needs to be repeatedly adjusted, increasing labor and time costs. Moreover, due to the diversity of processes, existing detection formulas are difficult to apply to defect detection in new processes, and adjusting the parameters of the detection formula requires a certain level of algorithmic expertise, thus placing high demands on users.

[0004] It should be noted that the information disclosed in the background section of this invention is intended only to enhance the understanding of the general background of this invention, and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to address the deficiencies in existing technologies by providing a method, system, electronic device, and storage medium for setting and optimizing detection formulas. The detection formula setting and optimization method provided by this invention, based on prior knowledge of detection result data and fully considering the coupling relationship between parameters, determines the strategy and parameter setting values ​​of the detection formula in one go. This not only improves the efficiency of the detection process but also enhances the detection accuracy of the detection formula.

[0006] To achieve the above objectives, the present invention provides a method for setting and optimizing a detection formula, comprising:

[0007] The first data sample is labeled to obtain the second data sample; wherein, the first data sample includes several detection result data; the second data sample includes the detection result data and a label corresponding to each detection result data;

[0008] Based on the second data sample, the data feature distribution information of the detected object is obtained;

[0009] A preset outlier statistical analysis strategy is used to perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information, and the detection formula is determined according to the preset outlier statistical analysis strategy.

[0010] Based on the defect distribution boundary information and the preset outlier statistical analysis strategy, the values ​​of the detection parameters of the detection formula are set or optimized through reverse derivation.

[0011] Optionally, the detection result data includes basic information and feature data information of the detection object; wherein, the feature data information includes the position information of the detection result on the detection object, and one or more of the following: the process flow information of the detection object, the grayscale information, shape information, and texture information of the data information of the detection result;

[0012] The step of labeling the first data sample to obtain the second data sample includes:

[0013] Obtain the basic information of the detection object corresponding to each detection result data in the first data sample;

[0014] For each detection result data, based on the basic information of the detection object and the location information of the detection result on the detection object, the original information corresponding to the detection result data on the detection object is obtained;

[0015] Based on the original information, determine whether the defect indicated by the data information of the detection result is a real defect. If so, mark the detection result data as real defect data; otherwise, mark the detection result data as noise data.

[0016] The second data sample is obtained based on all the test result data and the label corresponding to each test result data.

[0017] Optionally, the detection object includes a wafer; the basic information of the wafer includes the wafer's number, the number of dies it contains, and the basic information of each die; the basic information of the die includes the die number and image information of the die.

[0018] The step of obtaining the original information corresponding to the detection result data on the detection object based on the basic information of the detection object and the position information of the detection result on the detection object includes:

[0019] Based on the basic information of the Wafer, obtain the Die number of each Die of the Wafer and the basic information of each Die;

[0020] Based on the location information of the detection result on the die and the image information of the die, obtain the image information of the detection result corresponding to the detection result data on the die.

[0021] Optionally, obtaining the data feature distribution information of the detected object based on the second data sample includes:

[0022] A feature data axis and a segmentation data axis are determined, and a feature space is established based on the feature data axis and the segmentation data axis; wherein, the feature data axis represents the feature data information of the detection result data, and the segmentation data axis represents segmentation feature information; wherein, the segmentation feature information includes other feature data information besides that used for the feature data axis;

[0023] The second data sample is arranged according to the feature space to obtain the data feature distribution information of the detected object.

[0024] Optionally, the feature space includes one or more of the feature data axes and one or more of the segmentation data axes. Optionally, arranging the second data samples according to the feature space to obtain the data feature distribution information of the detected object includes:

[0025] Establish a rectangular coordinate system by taking the feature data axis as the horizontal axis and the segmentation data axis as the vertical axis;

[0026] In the Cartesian coordinate system, the second data samples are arranged according to the feature values ​​of the feature data information represented by the feature data axis in the horizontal direction and according to the feature values ​​of the feature data information represented by the segmentation data axis in the vertical direction to obtain a defect feature distribution map.

[0027] Optionally, the step of employing a preset outlier statistical analysis strategy to perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information includes:

[0028] If it is determined whether to automatically search for defect distribution boundary information, then the outlier statistical analysis model is trained according to the selected outlier statistical analysis model to obtain defect distribution boundary information; otherwise, the data segmentation method is used to perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information.

[0029] The step of training the outlier statistical analysis model includes: training the selected outlier statistical analysis model based on the detection result data and the data feature distribution information until the defect distribution boundary information of the detected object meets the first preset condition.

[0030] The step of performing outlier statistical analysis on the data feature distribution information using a data segmentation method includes: obtaining at least one first segmentation threshold on the feature data axis and / or the segmentation data axis based on the detection result data and the data feature distribution information; and obtaining the defect boundary information based on the first segmentation threshold until the obtained defect distribution boundary information of the detected object meets a second preset condition.

[0031] Optionally, the segmented data axis represents process flow information; the step of performing threshold segmentation on the feature data axis and / or the segmented data axis based on the detection result data and the data feature distribution information until the obtained defect distribution boundary information of the detected object meets the second preset condition includes:

[0032] Based on the data feature distribution information and the consistency of the distribution of detection result data labeled as true defect data and labeled as noise data, the first segmentation threshold of the segmented data axis is determined.

[0033] Based on the data feature distribution information and the consistency of the distribution of detection result data labeled as true defect data and labeled as noise data, a second segmentation threshold for the feature data axis is determined;

[0034] Based on the first segmentation threshold of the segmentation data axis and the second segmentation threshold of the feature data axis, the defect distribution boundary information of the detected object is obtained.

[0035] Optionally, the adoption of a preset outlier statistical analysis strategy further includes: an outlier statistical analysis strategy that combines data segmentation and model learning;

[0036] The outlier statistical analysis strategy combining data segmentation and model learning includes: obtaining at least one first segmentation threshold on the segmentation data axis for the detection result data labeled as true defects based on the data feature distribution information; and training the selected outlier statistical analysis model based on the first segmentation threshold and the data feature distribution information until the defect distribution boundary information of the detected object satisfies a third preset condition.

[0037] Optionally, the step of setting or optimizing the values ​​of the detection parameters of the detection formula by reverse derivation based on the defect distribution boundary information and the preset outlier statistical analysis strategy includes:

[0038] Based on the preset outlier statistical analysis strategy, determine the reverse derivation strategy;

[0039] Based on the reverse derivation strategy, determine the input data information of the reverse derivation strategy;

[0040] Based on the input data information, determine the data distribution model of the detection result data;

[0041] Based on the data distribution model and the defect distribution boundary information, the detection parameters of the detection formula are determined;

[0042] Based on the strategy of the detection formula and the input data information derived in reverse, the values ​​of the detection parameters of the detection formula are set or optimized.

[0043] Optionally, the preset outlier statistical analysis strategy is a data segmentation method;

[0044] According to the data segmentation method, the data distribution density of the detection result data of the detection object is used as the reverse derivation strategy.

[0045] Based on the reverse derivation strategy of the statistical data distribution density, all detection result data of the detection object are used as the input data information;

[0046] Based on all the detection results data, it is assumed that the feature values ​​of all the detection results data feature information in the feature space are divided into normal region, noisy region and true defect region; the normal region is the region where the data distribution density is greater than the first density threshold, the noisy region is the region where the data density is less than or equal to the first density threshold and greater than the second density threshold, and the true defect region is the region where the data density is less than or equal to the second density threshold.

[0047] Based on all detection result data and the labels of all detection result data, calculate the first density threshold and the second density threshold; wherein, the first density threshold is greater than the second density threshold;

[0048] The displacement parameters of the detection formula are calculated based on the first density threshold, the second density threshold, and the defect distribution boundary information.

[0049] Optionally, the preset outlier statistical analysis strategy is an outlier statistical analysis strategy based on a Gaussian model;

[0050] According to the outlier statistical analysis strategy based on Gaussian model, the Gaussian distribution of the detection result data of the detection object is used as the reverse derivation strategy, and Gaussian model detection is used as the strategy for detection formulation.

[0051] Based on the reverse derivation strategy of statistical Gaussian distribution, all detection result data of the detection object and the defect distribution boundary information are used as the input data information.

[0052] Based on all the detection result data, it is assumed that the feature values ​​of all the feature data information of the detection result data follow a Gaussian distribution in the feature space.

[0053] The parameters for the Gaussian model detection are determined based on the input data and the defect distribution boundary information.

[0054] Optionally, the preset outlier statistical analysis strategy is a machine learning outlier statistical analysis strategy;

[0055] Based on the outlier statistical analysis strategy of the machine learning, the density threshold and distance threshold of the detection result data of the detection object are used as the reverse derivation strategy, and the machine learning model is used as the detection formula strategy.

[0056] Based on the reverse derivation strategy of the density threshold and distance threshold of the detection result data of the detection object, the density and distance of the detection result data of the detection object are used as the input data information.

[0057] Based on all the detection results data and the defect boundary distribution information, the density parameters and distance parameters of the detection strategy of the machine learning model are derived in reverse.

[0058] Optionally, the method for setting and optimizing the detection formula further includes:

[0059] Based on the test formula and the values ​​of the test parameters of the test formula, defect analysis is performed on the object to be tested to obtain defect data information of the object to be tested.

[0060] To achieve the above objectives, the present invention also provides a detection formula setting and optimization device, wherein the detection parameter and adjustment device includes:

[0061] The true defect and noise labeling unit is configured to label a first data sample to obtain a second data sample; wherein, the first data sample includes several detection result data; the second data sample includes the detection result data and a label corresponding to each detection result data;

[0062] The feature distribution information acquisition unit is configured to obtain the data feature distribution information of the detection object based on the second data sample;

[0063] The defect distribution boundary acquisition unit is configured to perform outlier statistical analysis on the data feature distribution information using a preset outlier statistical analysis strategy to acquire defect distribution boundary information, and to determine the detection formula according to the preset outlier statistical analysis strategy.

[0064] The detection parameter setting and optimization unit is configured to determine or optimize the values ​​of the detection parameters of the detection formula by reverse derivation based on the defect distribution boundary information and the preset outlier statistical analysis strategy.

[0065] Optionally, the detection formula setting and optimization device further includes:

[0066] The test formula application unit is configured to perform defect analysis on the object to be tested based on the test formula and the values ​​of the test parameters of the test formula, and obtain defect data information of the object to be tested.

[0067] To achieve the above objectives, the present invention also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the detection formula setting and optimization method described above.

[0068] To achieve the above objectives, the present invention also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the detection formula setting and optimization method described above.

[0069] Compared with the prior art, the detection formula setting and optimization method, apparatus, electronic device and storage medium provided by the present invention have the following advantages:

[0070] The detection formula setting and optimization method provided by this invention firstly obtains a second data sample by labeling a first data sample; wherein, the first data sample includes several detection result data; the second data sample includes the detection result data and a label corresponding to each detection result data; then, based on the second data sample, the data feature distribution information of the detection object is obtained, and the detection formula is determined according to the preset outlier statistical analysis strategy; next, the preset outlier statistical analysis strategy is used to perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information; finally, based on the defect distribution boundary information and the preset outlier statistical analysis strategy, the detection parameters of the detection formula are determined or optimized through reverse derivation. Therefore, the detection formula setting and optimization method provided by this invention includes a first data sample comprising several detection result data, which includes auxiliary parameter tuning information (such as basic information and feature data of the detection object, including but not limited to the grayscale, shape, texture, etc. of the defects indicated by the detection results). Data annotation can distinguish between true defect data and noisy data, providing an important basis for subsequent effective use of historical information for data analysis and reasoning to obtain accurate prior knowledge, thereby improving the detection accuracy of the detection formula. Furthermore, the detection formula setting and optimization method provided by this invention derives the detection formula strategy and detection parameter values ​​based on defect distribution boundary information and the preset outlier statistical analysis strategy through reverse derivation. Thus, this invention can simultaneously deduce a set of detection parameters (i.e., simultaneously adjust all parameters) through reverse derivation, taking into account the coupling relationship between parameters, achieving rapid modeling of the detection formula; avoiding repeated parameter adjustments, significantly saving manpower and time costs; and, for new process defect detection, users do not need algorithmic knowledge to set or optimize the detection formula strategy and the detection parameter values ​​of the detection formula.

[0071] Since the detection formula setting and optimization device, electronic device and storage medium provided by the present invention belong to the same inventive concept as the detection parameters and adjustment method provided by the present invention, the detection formula setting and optimization device, electronic device and storage medium provided by the present invention have all the advantages of the detection formula setting and optimization method, which will not be elaborated here. Attached Figure Description

[0072] Figure 1 A schematic flowchart illustrating the detection formula setting and optimization method provided in one embodiment of the present invention;

[0073] Figure 2 This is a schematic diagram of a data sample annotation method provided in one embodiment of the present invention;

[0074] Figure 3This is a schematic diagram of one interface for marking defects in data samples according to an embodiment of the present invention;

[0075] Figure 4 This is an example diagram showing the distribution of detection result data in a two-dimensional feature space, illustrating one specific example of applying the present invention.

[0076] Figure 5 This is a schematic diagram illustrating the principle of outlier statistical analysis provided in one embodiment of the present invention;

[0077] Figure 6 A schematic diagram of defect distribution boundary information obtained by applying the outlier statistical analysis model provided by this invention;

[0078] Figure 7 for Figure 1 Detailed flowchart of step S400;

[0079] Figure 8 A specific example diagram illustrating the reverse derivation of the detection formula setting and optimization method provided by the present invention;

[0080] Figure 9 This is a schematic diagram of the data density distribution of one type of detection result data provided in an embodiment of the present invention;

[0081] Figure 10 A schematic diagram of the distribution of true defect data within the average gray level range of a standard segmentation axis provided in an embodiment of the present invention;

[0082] Figure 11(a) is an example of multiple test images provided in one embodiment of the present invention;

[0083] Figure 11(b) is an example of a mean plot generated from multiple test plots in Figure 11(a);

[0084] Figure 11(c) is an example of a standard deviation plot generated from multiple test plots in Figure 11(a);

[0085] Figure 11(d) is an enlarged example of one of the test images;

[0086] Figure 11(e) is a schematic diagram of the defect locations detected using machine learning recipes;

[0087] Figure 12 A schematic diagram of the grayscale dynamic threshold provided for applying this invention;

[0088] Figure 13 This diagram illustrates a comparison between the test results obtained using the test formula set and optimization method provided by the present invention and the test results obtained using the original test formula.

[0089] Figure 14This is a structural block diagram of the detection formula setting and optimization device according to one embodiment of the present invention;

[0090] Figure 15 This is a block diagram of an electronic device according to one embodiment of the present invention.

[0091] The reference numerals in the attached figures are as follows:

[0092] 1-Noise data, 2-True defect data, 3-Defect distribution boundary curve, segment_value1, segment_value2-First segmentation threshold, A, A1, A2-Pixels;

[0093] 100 - True Defect and Noise Marking Unit, 200 - Feature Distribution Information Acquisition Unit, 300 - Defect Distribution Boundary Acquisition Unit, 400 - Detection Parameter Setting and Optimization Unit, 500 - Detection Formula Application Unit;

[0094] 601 - Processor, 602 - Communication interface, 603 - Memory, 604 - Communication bus. Detailed Implementation

[0095] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, further illustrates the detection formula setting and optimization method, apparatus, electronic device, and storage medium proposed in this invention. The advantages and features of this invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, used only to facilitate and clearly illustrate the embodiments of this invention. Please refer to the drawings to make the objectives, features, and advantages of this invention more apparent and understandable. It should be understood that the structures, proportions, sizes, etc., depicted in the accompanying drawings are only for illustrative purposes and to enable those skilled in the art to understand and read them, and are not intended to limit the implementation conditions of this invention. Any modifications to the structure, changes in proportions, or adjustments to the size, provided that the effects and objectives achieved by this invention are the same or similar, should still fall within the scope of the technical content disclosed in this invention.

[0096] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0097] One embodiment of the present invention provides a method for setting and optimizing a detection formula; specifically, please refer to... Figure 1 The diagram illustrates a flowchart of a detection formula setting and optimization method provided by an embodiment of the present invention. Figure 1 As shown, the detection formula setting and optimization method includes the following steps:

[0098] S100: Label the first data sample to obtain the second data sample; wherein, the first data sample includes several detection result data; the second data sample includes the detection result data and a label corresponding to each detection result data;

[0099] S200: Based on the second data sample, obtain the data feature distribution information of the detected object;

[0100] S300: Using a preset outlier statistical analysis strategy, perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information, and determine the detection formula according to the preset outlier statistical analysis strategy;

[0101] S400: Based on the defect distribution boundary information and the preset outlier statistical analysis strategy, the values ​​of the detection parameters of the detection formula are set or optimized through reverse derivation.

[0102] Therefore, the detection formula setting and optimization method provided by this invention includes a first data sample comprising several detection result data points. These detection result data points include a large amount of auxiliary parameter tuning information (such as basic information and feature data of the detection object, including but not limited to the grayscale, shape, and texture information of the defects indicated by the detection results). Data annotation can distinguish between true defect data and noisy data, providing an important basis for subsequent effective use of historical information for data analysis and reasoning to obtain accurate prior knowledge, thereby improving the detection accuracy of the detection formula. Furthermore, the detection formula setting and optimization method provided by this invention derives the detection formula strategy and parameter settings based on defect distribution boundary information and the preset outlier statistical analysis strategy through reverse derivation. Thus, this invention can simultaneously deduce a set of detection parameters (i.e., simultaneously adjust all parameters) through reverse derivation, taking into account the coupling relationship between parameters, achieving rapid modeling of the detection process, avoiding repeated parameter adjustments, and significantly saving manpower and time costs. Moreover, for defect detection in new processes, users do not need to have algorithmic knowledge to set or optimize the detection formula strategy and detection parameter values.

[0103] It should be specifically noted that the detection result data refers to the historical detection result data of the object being detected. For example, when initially setting the strategy and detection parameters of the detection formula used in the defect detection process, the strategy and detection parameters of the detection formula can be randomly or manually selected to obtain a certain amount of defect detection data, which constitutes the detection result data (i.e., the first sample data). When optimizing the strategy and detection parameters of the detection formula, the detection result data (i.e., the first sample data) includes all or a portion of the historical detection data of the detection formula to be optimized. For ease of understanding and explanation, the detection result data described below refers to historical detection data of wafer defects. Obviously, this is not a limitation of the present invention. In other embodiments, the detection formula setting and optimization method provided by the present invention can also be adapted to other detection formulas for initial detection of wafer defects, which will not be exemplified one by one.

[0104] Preferably, in one preferred embodiment, the detection result data includes basic information and feature data information of the detected object; wherein, the feature data information includes the position information of the detection result on the detected object, and one or more of the following: process flow information of the detected object, grayscale information, shape information, and texture information of the detection result data information. As those skilled in the art will understand, it is obvious that the detection result data information also necessarily includes conclusion information (defect data or non-defect data) used to indicate the detection result. For ease of understanding, specific examples of the detection result data information will be described below in comparison with the image information of the detection result; here, the detection result data information will not be exemplified further. Therefore, the detection result data includes basic information and feature data information (such as grayscale, shape, and texture information of nuisance) and other auxiliary parameter tuning information of the detected object, and the subsequent defect distribution map drawing and parameter back-inference process are based on the detection result data. Thus, the detection formula setting and optimization method provided by the present invention can improve the detection accuracy of the detection formula.

[0105] Preferably, in one embodiment, please refer to Figure 2 It schematically illustrates the data sample annotation method flowchart. From Figure 2 As can be seen, in step S100, the step of labeling the first data sample to obtain the second data sample includes:

[0106] S110: Obtain the basic information of the detection object corresponding to each detection result data in the first data sample;

[0107] S120: For each detection result data, based on the basic information of the detection object and the position information of the detection result on the detection object, obtain the original information corresponding to the detection result data on the detection object;

[0108] S130: Based on the original information, determine whether the defect indicated by the data information of the detection result is a real defect. If yes, mark the detection result data as real defect data; otherwise, mark the detection result data as noise data.

[0109] S140: Based on all the detection result data and the label corresponding to each detection result data, the second data sample is obtained.

[0110] With this configuration, the detection formula setting and optimization method provided by the present invention can accurately distinguish between the real defect data and the nusiance data in the detection result data (historical data) by annotating the first data sample. This provides accurate prior knowledge for subsequent acquisition of data feature distribution information and further reverse deduction based on the defect distribution boundary information, thereby improving the detection accuracy of the detection formula.

[0111] It should be noted that those skilled in the art should understand that the feature data information refers to all the detection results of defect detection of the object being detected, including defect data and non-defect data.

[0112] As a preferred example of the detection formula setting and optimization method provided by the present invention, the following description uses a wafer as the detection object. Clearly, the first data sample is the historical detection result data of the wafer. More specifically, the basic information of the wafer includes the wafer's number, the number of dies it contains, and the basic information of each die; the basic information of the die includes the die number and image information. Correspondingly, in step S120, obtaining the original information corresponding to the detection result data on the detection object based on the basic information of the detection object and the location information of the defect on the detection object includes:

[0113] S121: Based on the basic information of the Wafer, obtain the Die number of each Die of the Wafer and the basic information of each Die;

[0114] S122: Based on the location information of the detection result on the die and the image information of the die, obtain the image information of the detection result corresponding to the detection result data on the die.

[0115] To facilitate a more accurate understanding of the present invention, the data information and image information of the detection results are explained below. The data information of the detection results includes a description of the image information of the detection results within the detection result data. The image information of the detection results is the original image corresponding to the data information of the detection results on the detected object. In other words, the data information of the detection results includes the data representation of the image information of the detection results. Taking a wafer as the detection object as an example: if the defect is a texture defect, the data information of the detection results records the texture features of the texture defect, such as the roughness of the texture, while the image information of the detection results is the original image corresponding to the texture defect. Therefore, based on the image information of the detection results, the detection result data corresponding to the image information of the detection results can be re-evaluated to determine whether it is genuine defect data or noise data.

[0116] Specifically, please see Figure 3 The diagram illustrates one type of interface for defect annotation of data samples according to an embodiment of the present invention. Figure 3As can be seen, the defect annotation interface has three main functional areas: a Wafer display window, a detection data list window, and a defect display area. Specifically, the Wafer display window graphically displays the basic information of the Wafer, including but not limited to the position of each die on the Wafer and the die number. Below the Wafer display window, the user can select the die number to be annotated. Based on the selected die number, the historical detection data results for the die corresponding to the selected die number will be refreshed in the detection data list window. Therefore, based on the list of detection result data in the detection data list window area, the user can select the detection result data one by one. The original information corresponding to the detection result data will be displayed in the defect display area (that is, the image information of the detection result is the image information indicating the position information of the detection result on the die). Thus, based on various characteristics of the original information (texture, size, curvature, shape, etc.), the defect indicated by the data information of the detection result can be further confirmed by manual review or machine review to determine whether it is a real defect. If it is, the detection result data is marked as real defect data (for example, the label of the detection result data in the detection data list window area is marked as real defect, and the value corresponding to the manual judgment of whether it is a real defect is set to yes); if not, the detection result data is marked as noisy data (for example, the label of the detection result data in the detection data list window area is marked as real defect, and the value corresponding to the manual judgment of whether it is a real defect is set to no). By repeating the above process, selecting each Die number in turn, and manually annotating each detection data result under the current Die, the annotation of the detection data results of the entire Wafer can be completed. In this way, the first data sample can be annotated, thereby obtaining the second data sample.

[0117] It should be noted that although the above description uses manual annotation as an example to illustrate the annotation method of the first data sample, this is clearly not a limitation of the present invention. In other embodiments, annotation can also be performed using methods such as machine learning, and the present invention does not limit this. Furthermore, as mentioned above, although the detection formula setting and optimization method provided by the present invention uses a wafer as an example of the detection object for illustration, as those skilled in the art will understand, this is only an exemplary illustration of a preferred embodiment and not a limitation of the present invention. In other embodiments, the detection object can also be other products besides wafers, including but not limited to lenses, displays, 3D printed products, etc., which will not be illustrated in detail here.

[0118] Preferably, in one exemplary embodiment, step S200, obtaining the data feature distribution information of the detected object based on the second data sample, includes:

[0119] S210: Determine the feature data axis and the segmentation data axis, and establish a feature space based on the feature data axis and the segmentation data axis; wherein, the feature data axis represents the feature data information of the detection result data, and the segmentation data axis represents segmentation feature information; wherein, the segmentation feature information includes other feature data information besides that used for the feature data axis;

[0120] S220: Arrange the second data samples according to the feature space to obtain the data feature distribution information of the detected object.

[0121] With this configuration, the detection formula setting and optimization method provided by the present invention arranges the second data sample through the feature space. The purpose is to make the distribution of the detection result data in the feature space show a certain trend, making the distinction between true defect data and noise data more obvious, so as to obtain defect distribution boundary information.

[0122] Preferably, the feature space includes one or more of the feature data axes and one or more of the segmentation data axes.

[0123] The detection formulation and optimization method provided by this invention includes a feature space that can include multiple feature data axes and multiple segmentation data axes, and the feature space can be a multi-dimensional feature space. For example, there are two feature data axes, one representing the grayscale information of the defect and the other representing the texture information of the defect; one segmentation data axis represents the shape information of the defect and the other represents the size of the defect. Therefore, the detection formulation and optimization method provided by this invention, by referencing more feature information of the defect, lays a good foundation for further improving the detection accuracy of the detection formulation. It should be noted that the above is merely an illustrative description and not a limitation of this invention. In practical applications, the feature data axes, the segmentation data axes, and their respective numbers should be reasonably selected according to actual needs.

[0124] Preferably, in one exemplary embodiment, step S220, arranging the second data samples according to the feature space to obtain the data feature distribution information of the detected object, includes:

[0125] S221: Establish a rectangular coordinate system by taking the feature data axis as the horizontal axis and the segmentation data axis as the vertical axis;

[0126] S222: In the rectangular coordinate system, the second data samples are arranged according to the feature value of the feature data information represented by the feature data axis in the horizontal direction and according to the feature value of the feature data information represented by the segmentation data axis in the vertical direction to obtain a defect feature distribution map.

[0127] Specifically, please see Figure 4 It schematically illustrates the distribution of detection results data in a two-dimensional feature space for one specific example. From Figure 4 As can be seen, this example uses a two-dimensional data feature distribution map where the horizontal axis represents the feature data axis and the vertical axis represents the segmentation data axis. That is, the horizontal coordinate of each point in the coordinate system represents the magnitude of the feature value, and the vertical coordinate represents the magnitude of the corresponding segmentation feature value. In this way, the feature values ​​of all the detection result data constitute the entire feature distribution map.

[0128] It should be noted that, as mentioned above, although the example uses a two-dimensional feature space distribution, in practical applications, the feature data axis and the segmentation data axis can be multi-dimensional. That is, the segmentation data axis can select multiple segmentation values ​​to divide the detection result data (i.e., the second sample data) into several different feature distributions.

[0129] Furthermore, this invention does not limit the specific method for selecting the feature space. In one embodiment, a feature selection algorithm can be used to automatically select the feature data axis and the segmentation data axis; in other embodiments, the feature data axis and the segmentation data axis can also be selected manually, and this invention does not impose any limitations on this. More specifically, the feature data axis can represent information such as color, texture, shape, and size, and the segmentation data axis can be information such as a trained mean plot.

[0130] Furthermore, as a preferred embodiment, the criterion for selecting the feature space is: the segmentation data axis can effectively distinguish different process regions; the feature data axis can clearly differentiate between true defect data and noise data (noise points); the ultimate goal is to make the distribution of the detection result data in the feature space exhibit a certain trend, making the distinction between true defects and noise points more obvious. For example, for wafer defect detection result data, if using the shape in the feature data information as the feature data axis is more effective in distinguishing between true defects and noise data in the feature space than using the texture in the feature data information as the feature data axis, then the shape in the feature data information is used as the feature data axis, rather than the texture in the feature data information. Understandably, the shape in the feature data information is no longer used as the segmentation data axis.

[0131] Preferably, in one preferred embodiment, please refer to Figure 5 The diagram illustrates a flowchart of a detection formula setting and optimization method provided by an embodiment of the present invention. Figure 5 As can be seen, in step S300, the step of using a preset outlier statistical analysis strategy to perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information includes:

[0132] If it is determined whether to automatically search for defect distribution boundary information, then the outlier statistical analysis model is trained according to the selected outlier statistical analysis model to obtain defect distribution boundary information; otherwise, the data segmentation method is used to perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information.

[0133] Specifically, please see Figure 6 ,in, Figure 6 A schematic diagram of the defect distribution boundary information obtained by applying the outlier statistical analysis model provided by this invention. Figure 6 In the diagram, feature1 represents the segmentation data axis, and feature2 represents the feature data axis. From... Figure 6 As can be seen, in this example, the defect distribution boundary information 3 is a curve. Therefore, the detection formula setting and optimization method provided by this invention, by determining the preset outlier statistical analysis strategy based on the detection result data and the data feature distribution information, and performing outlier statistical analysis on the data feature distribution information according to the determined preset outlier statistical analysis strategy to obtain defect distribution boundary information, enables the defect distribution boundary information to effectively separate true defect data 2 and noisy data 1. That is, the defect distribution boundary information can minimize over-detection problems without producing missed defects, thereby filtering out more noise data. This ensures that the detection formula subsequently determined based on the defect distribution boundary information does not result in missed or over-detection, thus improving the defect detection accuracy of the detection process.

[0134] It is important to note that even with the same second sample data and the same feature space, different outlier statistical analysis strategies may yield different defect distribution boundary information. Therefore, subsequent reverse derivation and detection formulation strategies are closely related to the outlier statistical analysis strategy. Figure 6 Using the same second sample data, such as data segmentation, the shape of the defect distribution boundary information is similar to... Figure 6 It is completely different from China; please refer to the description below for details. To avoid redundancy, it will not be elaborated here.

[0135] For ease of understanding and explanation, the following explanations will use two-dimensional data distribution as an example. First, the outlier statistical analysis model will be explained in detail, and then the data segmentation method will be explained.

[0136] Specifically, training the outlier statistical analysis model includes: training the selected outlier statistical analysis model based on the detection result data and the data feature distribution information until the defect distribution boundary information of the detected object meets the first preset condition.

[0137] More specifically, as those skilled in the art will understand, an outlier statistical analysis model can be selected based on a comprehensive analysis of the detection result data and the data feature distribution information. This outlier analysis statistical model includes, but is not limited to, statistical outlier algorithms (such as the 3σ principle), distance and proximity-based clustering algorithms (such as K-means), density-based outlier algorithms (such as DBSCAN), and tree-based outlier analysis algorithms (such as isolated forests). It is particularly important to note that the choice of algorithm model is crucial; different algorithm models imply different outlier boundary shapes. An optimal algorithm model ensures that the training of the dataset avoids both underfitting and overfitting. For example, if the distribution of the second sample data in the feature space is closer to a normal distribution, then a statistical outlier analysis statistical model (such as the 3σ principle) is preferred. Furthermore, if the distribution of the second sample data in the feature space shows that the distance between true defect data and noisy data is relatively short, while the distance between defect data and noisy data is relatively long, then a distance and proximity-based clustering algorithm is preferred. Those skilled in the art should be able to apply this principle to other situations, so further details will not be elaborated here.

[0138] Furthermore, those skilled in the art should understand that the purpose of the outlier analysis statistical model is to find the optimal boundary result. After determining the outlier analysis statistical model, the selected outlier analysis statistical model should be trained using the second sample data. Through continuous learning and target optimization processes, the model training result can find the optimal inflection point of the segmentation data axis and distinguish true defects from noisy data (interference noise points) based on the feature data axis. Thus, after the outlier analysis statistical model training is completed, a boundary result (i.e., defect distribution boundary information) is obtained. Please refer to [link to relevant documentation]. Figure 6 ,like Figure 6 As shown, the defect distribution boundary curve 3 (i.e., defect distribution boundary information) can effectively separate defect data and noise data, ensuring that the detection results do not miss any detections or produce over-detections. That is, the first preset condition is that the defect distribution boundary information can distinguish between the detection result data labeled as true defect data and the detection result data labeled as noise data in the second sample.

[0139] Furthermore, the outlier statistical analysis of the data feature distribution information using the data segmentation method includes: obtaining at least one first segmentation threshold on the feature data axis and / or the segmentation data axis based on the detection result data and the data feature distribution information; and obtaining the defect boundary information based on the first segmentation threshold until the obtained defect distribution boundary information of the detected object meets the second preset condition.

[0140] In one preferred embodiment, the data segmentation method includes manually segmenting the feature space to obtain the first segmentation threshold. As those skilled in the art will understand, this invention does not limit the specific implementation of the data segmentation method; in other embodiments, the first segmentation threshold can also be obtained using a data segmentation algorithm.

[0141] To facilitate understanding and explanation, the data segmentation method will be illustrated below using two-dimensional data distribution and manual segmentation as examples:

[0142] S321: Based on the data feature distribution information and the consistency of the distribution of detection result data labeled as true defect data and labeled as noise data, determine the first segmentation threshold of the segmented data axis.

[0143] S322: Based on the data feature distribution information and the consistency of the distribution of detection result data labeled as true defect data and labeled as noise data, determine the second segmentation threshold of the feature data axis;

[0144] S323: Based on the first segmentation threshold of the segmentation data axis and the second segmentation threshold of the feature data axis, the defect distribution boundary information of the detected object is obtained.

[0145] Specifically, in step S321, the data feature distribution information is used as input to segment the data axis in this feature distribution map. The segmentation criterion is the consistency of the distribution of the detection result data. Data with a consistent distribution are grouped into clusters, and the segmentation values ​​between clusters are found to distinguish data from different processes. The consistent distribution includes the distribution pattern of the feature data information of the detection result data, including but not limited to the distribution density in the feature space and the relative positional relationship of spatial points. This serves as the basis for determining the segmentation axis and feature axis thresholds. For example, in one example, two first segmentation thresholds, segment_value1 and segment_value2, are set.

[0146] Correspondingly, in step S322, a second segmentation threshold is determined for the feature data axis in the feature distribution. Since the defect data points have already been labeled in the feature distribution, the principle for determining the second segmentation threshold is to separate the noisy data and the true defect data as far apart as possible. This ensures that the detection results do not miss any detections while minimizing the occurrence of over-detections. That is, the second preset condition is preferably that the defect boundary information can separate the true defect data and the noisy data.

[0147] Therefore, after determining the first segmentation threshold for the segmented data axis and the second segmentation threshold for the feature data axis, the defect boundary information for outlier statistical analysis can be obtained. The following figure, still using a two-dimensional feature data distribution as an example, illustrates the manually segmented defect distribution boundary information. Two first segmentation thresholds, segment_value1 and segment_value2, are used on the segmentation axis to segment the detection result data, dividing all the detection result data into three different distributions. Within each segmentation threshold interval, three different second segmentation thresholds are used on the feature data axis to distinguish between true defects and noisy data, obtaining the final defect distribution boundary information. That is, the defect distribution boundary information includes two straight lines parallel to the feature data axis feature1 formed by two first segmentation thresholds segment_value1 and segment_value2, and a first line segment located between the feature data axis feature1, the first segmentation thresholds segment_value1 and segment_value2, intersecting the feature data axis feature1 and the first segmentation thresholds segment_value1, a second line segment intersecting the first segmentation thresholds segment_value1 and segment_value2, and a third straight line intersecting the first segmentation thresholds segment_value2 and extending upward along the segmentation data axis feature2.

[0148] Preferably, in one exemplary embodiment, the pre-defined outlier statistical analysis strategy further includes an outlier statistical analysis strategy combining data segmentation and model learning. The outlier statistical analysis strategy combining data segmentation and model learning includes: obtaining at least one first segmentation threshold on the segmentation data axis for the detection result data labeled as true defects based on the data feature distribution information; and training the selected outlier statistical analysis model based on the first segmentation threshold and the data feature distribution information until the defect distribution boundary information of the detected object satisfies a third pre-defined condition.

[0149] With this configuration, the detection recipe setting and optimization method provided by the present invention can further reduce the uncertainty of machine learning model training by using an outlier statistical analysis strategy that combines data segmentation and model learning when acquiring outlier distribution boundary information. This gives the input of the machine learning model certain constraints, using the results of manual segmentation as constraints, thereby further improving the efficiency of acquiring defect boundary distribution information.

[0150] The third preset condition is preferably one that ensures no missed detections in the detection results while minimizing over-detection. Specifically, the second preset condition is preferably one where the defect boundary information can separate the true defect data from the noisy data, or the training times of the outlier statistical analysis model reach a preset value.

[0151] As those skilled in the art will understand, unlike the data segmentation method, the defect distribution boundary information obtained by the outlier statistical analysis strategy combining data segmentation and model learning differs from the defect distribution boundary information obtained by the data segmentation method described above. The defect distribution boundary information obtained by the outlier statistical analysis strategy combining data segmentation and model learning includes two straight lines parallel to the feature data axis feature1, formed by two first segmentation thresholds segment_value1 and segment_value2, and closed curves enclosing the true defect data, located in three intervals formed by the feature data axis feature1, the first segmentation thresholds segment_value1, and segment_value2, respectively. Because different outlier statistical analysis strategies are used, the obtained defect boundary distribution information is drastically different. However, it is obvious that regardless of the outlier statistical analysis strategy used, the obtained defect boundary distribution information can accurately distinguish between true defect data and noisy data in the detection result data. As mentioned above, based on this, the present invention does not limit the specific implementation method of the outlier statistical analysis strategy.

[0152] Furthermore, for details regarding the data segmentation method and model learning in the outlier statistical analysis strategy that combines data segmentation and model learning, please refer to the detailed explanation of the data segmentation method and outlier statistical analysis model above. To avoid redundancy, they will not be elaborated upon here.

[0153] Preferably, in one exemplary embodiment, please refer to Figure 7 It schematically gives Figure 1 A detailed flowchart of step S400. From Figure 7 As can be seen, in step S400, the step of determining the values ​​of the detection parameters for setting or optimizing the detection formula through reverse derivation based on the defect distribution boundary information and the preset outlier statistical analysis strategy includes:

[0154] S410: Determine the reverse derivation strategy based on the preset outlier statistical analysis strategy;

[0155] S420: Determine the input data information of the reverse derivation strategy according to the reverse derivation strategy;

[0156] S430: Determine the data distribution model of the detection result data based on the input data information;

[0157] S440: Determine the detection parameters of the detection formula based on the data distribution model and the defect distribution boundary information;

[0158] S450: Based on the strategy of the detection formula and the input data information derived in reverse, set or optimize the values ​​of the detection parameters of the detection formula.

[0159] Therefore, compared with the existing technology where forward parameter setting involves adjusting parameters based on the feedback from forward parameter tuning (which may only adjust one or two detection parameters), the detection formula setting and optimization method provided by this invention uses reverse derivation to determine the detection formula strategy and infers all parameter settings (key parameters, such as data density, data sparsity distance, and / or tolerance range) of the detection formula based on defect boundary distribution information. It also considers the coupling relationship between parameters in the detection process, thus avoiding repeated parameter tuning. Moreover, based on user annotation results, the parameter tuning process can automatically deduce a relatively accurate set of detection process parameters without prior knowledge, adjusting all detection parameters to the optimal level at once. This improves both the efficiency of detection process parameter tuning and the detection accuracy of the detection formula.

[0160] For more details, please see Figure 8 The diagram illustrates a specific example of reverse derivation using the detection formula setting and optimization method provided by this invention. From... Figure 8 It can be seen that the outlier statistical analysis strategy, the reverse derivation strategy, and the parameter settings of the detection process in the detection formula setting and optimization method provided by this invention are closely related: that is, the core of the reverse derivation strategy and the detection formula strategy is consistent with the core of the outlier statistical analysis strategy for obtaining the defect boundary distribution information. For example, if the outlier segmentation method is used as the outlier statistical analysis strategy, then the basic principles of the reverse derivation and detection formula strategies should also be consistent with the basic principles of the outlier segmentation method.

[0161] To facilitate understanding of this invention, the following examples illustrate the process of reverse derivation to obtain the parameter settings of the detection formula, using data segmentation as an outlier statistical analysis strategy, Gaussian model-based outlier statistical analysis strategy, and machine learning-based outlier statistical analysis strategy.

[0162] I. Data partitioning method: reverse derivation of new data flow and parameter settings.

[0163] Before explaining the specific steps for reverse-engineering the basic principles of the data segmentation method to obtain the detection formula and parameter settings, the core idea of ​​this method is explained as follows:

[0164] To facilitate understanding of this invention, please refer to... Figure 9 This diagram illustrates the data density distribution of one type of detection result data provided in the first embodiment of this method. The basic idea of ​​this method is to define the region where the density of points (feature values ​​of the detection result data) in the feature distribution map is greater than a first threshold as a normal region, that is, to represent the normal region as a function related to data density. Therefore, all data points (feature values ​​of the detection result data) with a data density greater than the first threshold are considered normal, and the data density is one of the detection parameters that needs to be inferred in reverse. Further, the region where the data density is less than or equal to the first threshold and greater than a second threshold is defined as a nuisance region. The detection parameter of the nuisance region indicates that the detection result data in this region contains noise, and this noise is an acceptable error (i.e., a nuisance region is generated due to process errors and noise effects), and does not belong to defective data. That is, the nuisance region is considered to be described by adding a tolerance value (displacement parameter) to the normal region, expressed by the following formula:

[0165] nuisance_threshold = f1(data_density) (1)

[0166] Since the true defect data has already been labeled in the outlier statistical analysis (i.e., the boundaries between the noisy region and the true defect region have been obtained), the displacement parameters (tolerance values) can be inferred backward from the defect boundary distribution information. Regions with data density less than or equal to the second threshold (i.e., outside the noisy region) are defined as true defect regions, specifically, this can be expressed by the following formula:

[0167] boundary threshold = f2(inspection_data) (2)

[0168] defect_threshold = f3(boundary threshold) (3)

[0169] offset_parameter = abs (defect_threshold - nuisance_threshold) (4)

[0170] In the formula, boundary threshold is the defect distribution boundary result obtained by the outlier statistical analysis algorithm, defect_threshold is the function related to the defect distribution boundary boundary threshold, and finally the displacement parameter offset_parameter can be calculated using defect_threshold and nuisance_threshold.

[0171] Based on the above analysis, as a preferred embodiment, if the preset outlier statistical analysis strategy is a data segmentation method, the displacement parameters of the detection formula are obtained through the following steps:

[0172] Step A1: Based on the data segmentation method, the data distribution density of the detection result data of the detection object is used as the reverse derivation strategy.

[0173] Step A2: Based on the reverse derivation strategy of the statistical data distribution density, use all the detection result data of the detection object as the input data information.

[0174] Step A3: Based on all the detection result data, assume that the feature values ​​of all the detection result data feature information in the feature space are divided into normal region, noisy region and true defect region; the normal region is the region where the data distribution density is greater than the first density threshold, the noisy region is the region where the data density is less than or equal to the first density threshold and greater than the second density threshold, and the true defect region is the region where the data density is less than or equal to the second density threshold.

[0175] Step A4: Calculate the first density threshold and the second density threshold based on all detection result data and the labels of all detection result data; wherein, the first density threshold is greater than the second density threshold;

[0176] Step A5: Calculate the displacement parameters of the detection formula based on the first density threshold, the second density threshold, and the defect distribution boundary information.

[0177] More specifically, in order to better understand the present invention, the following section will take wafer macro-defect detection as an example to explain in detail the parameter settings for obtaining defect boundary distribution information by data segmentation and reverse derivation of the detection process.

[0178] See Figure 10This schematically illustrates the distribution of true defects within the average grayscale range of the standard segmentation axis provided in the first embodiment of this diagram. Figure 10 As shown, it is assumed that defects exist within each grayscale level range of the standard segmentation axis (i.e., the segmentation data axis, corresponding to the vertical axis Feature2 in the figure), and the true defect data and noisy data are labeled. The standard segmentation axis is defined by N (N can be set according to actual needs, for example...). Figure 10 (N=10, this invention does not limit this) The average image generated by statistically analyzing N standard defect-free process data images is used as the final result, i.e., the average pixel grayscale value of the corresponding pixels in the N standard images is used. For details, please refer to... Figures 11(a)-11(c) as well as Figure 12 In this embodiment, Figure 11(a) is an example of multiple test charts provided by one embodiment of the present invention, Figure 11(b) is an example of a mean chart generated from the multiple test charts in Figure 11(a), and Figure 11(c) is an example of a standard deviation chart generated from the multiple test charts in Figure 11(a). Figure 12 This is a schematic diagram illustrating the application of the grayscale dynamic threshold provided by the present invention. In the diagram, pixel A is a pixel in the test image, and pixels A1 and A2 are the corresponding pixels of pixel A in the mean image and standard deviation image, respectively.

[0179] a. Selecting a sample: Select a sample (as shown in Figure 11(a)) and obtain the mean plot and standard deviation plot based on the N test plots (training).

[0180] b. Determine the feature data axis and the segmentation data axis, see [link / reference] Figure 10 feature1 is Figure 10 The feature data axis in the image, fearture2 is... Figure 10 The segmentation data axis is used. Based on the grayscale values ​​of the test images and the grayscale values ​​of the segmentation data axis, the value of each test image on the feature data axis (feature1) is calculated, as shown in the following formula:

[0181] feature1 = test - mean (4)

[0182] In the formula, feature1 is Figure 10 In the feature data axis, test is the grayscale value of the test image, and mean is the grayscale value of the average value of N test images.

[0183] As mentioned earlier, the segmented data axis feature2 is obtained by the following formula:

[0184] feature2 = mean (5)

[0185] In the formula, mean is the average grayscale value of the corresponding pixel in the standard image.

[0186] c. Assume that the defect distribution boundary information (threshold) is represented by the following formula:

[0187] defect_threshold=mean+ / -(sigma*std+gray) (6)

[0188] In the formula, mean is the average grayscale value of the corresponding pixel in the standard image, std is the standard deviation of one pixel in the test image, sigma is the coefficient of the standard deviation, and gray is the parameter to be solved. The dynamic threshold gray is equivalent to the offset parameter mentioned above, which can be defined as any curve. It has the following relationship with the average grayscale value mean:

[0189] gray=b+a1*mean+a2*mean^2+a3*mean^3+……+am*mean^m (7)

[0190] When only the first two terms of the polynomial are taken, the dynamic threshold gray=b+a1*mean is a straight line; when subsequent terms of the polynomial are taken, it becomes a curve. Substituting the corresponding values ​​of multiple points into equation (6), we get:

[0191]

[0192] Therefore, the problem of detecting parameters by reverse inference is transformed into the problem of finding the optimal solution of the above system of equations using the least squares method, where the value of sigma is the coefficient of the variance std in the boundary threshold assumption formula, and [b a1 a2 … an] are all the coefficients in the piecewise curve to be fitted above.

[0193] d. Solve for the coefficients in the polynomial.

[0194] Solving the system of equations yields the values ​​of sigma and the coefficients [b a1 a2 … an], thus resolving all the parameters involved in the algorithm. See the following equation:

[0195]

[0196]

[0197] The above system of equations can be transformed into matrix form as follows:

[0198] A x=b

[0199] A'A x=A'b

[0200] x=(A'A)^(-1)*(A'b)

[0201] Therefore, x is the final solution, which can be obtained as a vector through the matrix operations described above:

[0202]

[0203] Based on the above vectors, the coefficient sigma of the standard deviation in the detection process, as well as several coefficients required for the dynamic threshold curve, can be obtained. Therefore, the dynamic threshold gray curve can also be derived. Thus, the true defect threshold for each pixel in the test image can be obtained using the following formula in the detection process:

[0204] defect_threshold=std*sigma+gray (6)

[0205] Pixels that are greater than the threshold defect_threshold are considered normal pixels, while pixels that are less than or equal to the threshold defect_threshold are considered defective pixels.

[0206] Specifically, please refer to Figures 11(d) and 11(e), where Figure 11(d) is an enlarged example of one of the test images, and Figure 11(e) is a schematic diagram of the defect location detected using a machine learning algorithm. By comparing Figures 11(d) and 11(e), it is easy to see that the detection formula obtained using the detection formula setting and optimization method provided by this invention can accurately detect the true defects of the object to be detected.

[0207] II. Outlier Statistical Analysis Strategy Based on Gaussian Model: Reverse Derivation of New Data Flow and Parameter Settings

[0208] To facilitate understanding of this invention, before specifically describing the outlier-based outverse derivation method for obtaining new data flows and parameter settings provided by this invention, the core idea of ​​the outlier statistical analysis strategy based on Gaussian models for back-deriving new data flows and parameter settings will be explained first. The basic principle of this method is to assume that the distribution of all data points (detection result data) in the feature distribution map follows a Gaussian distribution. Then, based on the defect boundary distribution information in outlier statistical analysis, the mean, variance, and variance coefficients required in the detection model (detection flow strategy) are back-derived to obtain the relevant parameters needed for detection using the Gaussian model. Similar to the data segmentation method, the process of back-deriving new data flows and parameter settings, based on the outlier statistical analysis strategy of Gaussian models, includes the following steps:

[0209] Step B1: The preset outlier statistical analysis strategy is an outlier statistical analysis strategy based on Gaussian model;

[0210] Step B2: According to the Gaussian model-based outlier statistical analysis strategy, the Gaussian distribution of the detection result data of the detection object is used as the reverse derivation strategy, and the Gaussian model detection is used as the strategy for the detection formula.

[0211] Step B3: Based on the reverse derivation strategy of statistical Gaussian distribution, use all the detection result data of the detection object as the input data information and the defect distribution boundary information as the input data information;

[0212] Step B4: Based on all the detection result data, assume that the feature values ​​of all the feature data information in the feature space follow a Gaussian distribution.

[0213] Step B5: Determine the parameters of the Gaussian model detection based on the input data information and the defect distribution boundary information.

[0214] More specifically, this can be expressed by the following functional relationships:

[0215] boundary threshold=f2(inspection_data) (6)

[0216] μ = f4(inspection_data) (7)

[0217] ∑=f5(inspection_data,,μ) (8)

[0218] ∏=f6(boundary threshold, inspection_data, μ,, ∑) (9)

[0219] In the formula, `boundary threshold` represents the defect boundary distribution result obtained by the outlier algorithm, and this boundary matrix `boundary threshold` can be obtained. The mean μ can be obtained from the detection result data; it is obtained by averaging the gray values ​​of the current detection data image. The variance ∑ is calculated by summing the squares of the subtractions between the gray values ​​of the pixels in the image to be detected and the mean μ, and then averaging the results. The weight ∏ can be expressed as a function relating `boundary threshold`, `inspection_data`, μ, and ∑; it is represented by the coefficient of the variance ∑, according to the following formula:

[0220] μ+∏*∑=boundary threshold (10)

[0221] In the above equation, since μ, variance ∑ and boundary threshold have already been calculated, the weight ∏ can be obtained by solving the equation.

[0222] III. Outlier Statistical Analysis Strategies Based on Machine Learning: Back-Deriving New Data Flow and Parameter Settings

[0223] Preferably, in one exemplary embodiment, the machine learning-based outlier statistical analysis strategy, which reverse-engineers new data flows and parameter settings, includes the following steps:

[0224] Step C1: The preset outlier statistical analysis strategy is a machine learning outlier statistical analysis strategy;

[0225] Step C2: Based on the outlier statistical analysis strategy of the machine learning, the density threshold and distance threshold of the detection result data of the detection object are used as the reverse derivation strategy, and the machine learning model is used as the strategy for the detection formula.

[0226] Step C3: Based on the reverse derivation strategy of the density threshold and distance threshold of the detection result data of the detection object, the density and distance of the detection result data of the detection object are used as the input data information;

[0227] Step C4: Based on all detection result data and the defect boundary distribution information, deduce the density parameters and distance parameters of the detection strategy of the machine learning model in reverse.

[0228] As those skilled in the art will understand, since machine learning models require the specification of multiple parameters, the determination of these parameters directly affects the detection accuracy of outlier statistical analysis algorithms based on machine learning. Examples include the initial cluster centers in the k-means algorithm, and the neighborhood and quantity thresholds in the DBSCAN algorithm. Therefore, by using the defect boundary distribution information (results) from outlier statistical analysis to infer these machine learning parameters, a machine learning model with prior knowledge can be obtained, thereby improving the model's detection accuracy. Specifically, this can be achieved through the following formulas:

[0229] boundary threshold=f7(inspection_data)

[0230] density parameters=f8(boundary threshold,inspection_data)

[0231] distance parameters=f9(boundary threshold,inspection_data)

[0232] In the formula, the boundary threshold represents the defect boundary distribution information obtained by the outlier algorithm, which is related to the detection result data and has already been obtained in the defect boundary analysis process. Two important parameters for the distance-density-based clustering algorithm are density parameters and distance parameters. These parameters are derived from the detection result data and the boundary matrix. By inversely calculating the distance and density parameters, defects can be detected precisely because they are located outside the preset threshold; while normal pixels, located within the higher density threshold range, are filtered out, thereby improving detection accuracy.

[0233] Preferably, in one exemplary embodiment, please continue to refer to Figure 1 ,from Figure 1 It can be seen that the method for setting and optimizing the detection formula also includes:

[0234] S500: Based on the test formula and the values ​​of the test parameters of the test formula, perform defect analysis on the object to be tested to obtain defect data information of the object to be tested.

[0235] Please see Figure 13 It schematically illustrates a comparison between the test results obtained using the test formula setting and optimization method proposed in this invention and the test results obtained using the original test process. Figure 13 It can be seen that the detection process strategy and parameter settings obtained by reverse derivation of this invention are used in the detection process. Nuisance noise data is filtered out, and true defect data (defect data) is retained. The correctness of the results can be intuitively verified by the distribution of the detection result data in the feature space.

[0236] In summary, the detection formula setting and optimization method provided by this invention includes a first data sample comprising several detection result data points. These detection result data points include a large amount of auxiliary parameter tuning information. Through data annotation, it provides an important basis for subsequent effective use of historical information for data analysis and reasoning, thereby obtaining accurate prior knowledge and improving the detection accuracy of the detection formula. Furthermore, the detection formula setting and optimization method provided by this invention derives the detection formula strategy and parameter settings based on defect distribution boundary information and the preset outlier statistical analysis strategy through reverse derivation. Therefore, this invention can simultaneously deduce a set of detection parameters (and adjust all parameters simultaneously) through reverse derivation, taking into account the coupling relationships between parameters, thus achieving rapid modeling of the detection process; avoiding repeated parameter adjustments, significantly saving manpower and time costs; and, for defect detection in new processes, users do not need to have algorithmic knowledge to determine the detection process strategy and parameter settings.

[0237] Another embodiment of the present invention provides a detection formula setting and optimization device; for details, please refer to [link to relevant documentation]. Figure 14 The diagram schematically illustrates the structural block diagram of the detection formula setting and optimization device provided in this embodiment. From... Figure 14 As can be seen, the detection formula setting and optimization device provided in this embodiment includes: a true defect and noise marking unit 100, a feature distribution information acquisition unit 200, a defect distribution boundary acquisition unit 300, and a detection parameter setting and optimization unit 400.

[0238] Specifically, the true defect and noise labeling unit 100 is configured to label the first data sample to obtain a second data sample; wherein the first data sample includes several detection result data; the second data sample includes the detection result data and a label corresponding to each detection result data. The feature distribution information acquisition unit 200 is configured to obtain the data feature distribution information of the detection object based on the second data sample. The defect distribution boundary acquisition unit 300 is configured to perform outlier statistical analysis on the data feature distribution information using a preset outlier statistical analysis strategy to obtain defect distribution boundary information, and to determine the detection formula based on the preset outlier statistical analysis strategy. The detection parameter setting and optimization unit 400 is configured to set or optimize the values ​​of the detection parameters of the detection formula by reverse derivation based on the defect distribution boundary information and the preset outlier statistical analysis strategy.

[0239] Preferably, as one exemplary embodiment, the detection formula setting and optimization device further includes a detection formula application unit 500. Specifically, the detection formula application unit 500 is configured to perform defect analysis on the object to be tested based on the detection formula and the values ​​of the detection parameters of the detection formula, and obtain defect data information of the object to be tested.

[0240] Since the detection formula setting and optimization device provided by this invention is similar in basic principle to the detection formula setting and optimization methods provided in the above embodiments, the specific details of the above-described detection formula setting and optimization device embodiments are described in a relatively brief manner to avoid redundancy. For detailed information, please refer to the detailed description of the detection formula setting and optimization methods above. Furthermore, since the detection formula setting and optimization device provided by this invention and the detection formula setting and optimization methods provided in the above embodiments belong to the same inventive concept, the detection formula setting and optimization device provided by this invention has at least the same beneficial effects as the aforementioned detection formula setting and optimization methods. Please refer to the relevant content in the detection formula setting and optimization methods above; therefore, it will not be described in detail here. In addition, since the detection formula setting and optimization device in this invention and the detection formula setting and optimization methods described above belong to the same inventive concept, the description of the detection formula setting and optimization device in this article is relatively simple. For details on how it works, please refer to the relevant content in the detection formula setting and optimization methods above; therefore, it will not be described in detail here.

[0241] Based on the same inventive concept, the present invention also provides an electronic device, please refer to... Figure 15 The diagram illustrates a block structure of an electronic device according to an embodiment of the present invention. Figure 15 As shown, the electronic device includes a processor 601 and a memory 603. The memory 603 stores a computer program. When the computer program is executed by the processor 601, it implements the detection formula setting and optimization method described above. Since the electronic device provided by this invention and the detection formula setting and optimization method described above belong to the same inventive concept, it has all the advantages of the detection formula setting and optimization method described above, and therefore will not be described in detail here.

[0242] like Figure 15As shown, the electronic device also includes a communication interface 602 and a communication bus 604, wherein the processor 601, the communication interface 602, and the memory 603 communicate with each other via the communication bus 604. The communication bus 604 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus 604 can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is used to represent it in the figure, but this does not indicate that there is only one bus or one type of bus. The communication interface 602 is used for communication between the aforementioned electronic device and other devices.

[0243] The processor 601 referred to in this invention can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 601 is the control center of the electronic device, connecting various parts of the entire electronic device through various interfaces and lines.

[0244] The memory 603 can be used to store the computer program. The processor 601 implements various functions of the electronic device by running or executing the computer program stored in the memory 603 and calling the data stored in the memory 603.

[0245] The memory 603 may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0246] This invention also provides a readable storage medium storing a computer program, which, when executed by a processor, can implement the detection formula setting and optimization method described above. Since the readable storage medium provided by this invention and the detection formula setting and optimization method described above belong to the same inventive concept, it possesses all the advantages of the aforementioned detection formula setting and optimization method, and therefore will not be elaborated further.

[0247] The readable storage medium of embodiments of the present invention can be any combination of one or more computer-readable media. The readable medium can be a computer-readable signal medium or a computer-readable storage medium. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections having one or more wires, portable computer hard disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus, or device.

[0248] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0249] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0250] In summary, compared with existing technologies, the detection formula setting and optimization method, apparatus, electronic device, and storage medium provided by this invention have the following advantages: The first data sample includes several detection result data, which includes auxiliary parameter tuning information (such as basic information and feature data information of the detection object, including but not limited to the grayscale, shape, texture, etc. of the defects indicated by the detection results). Data annotation can distinguish between true defect data and noisy data, providing an important basis for subsequent effective use of historical information for data analysis and reasoning to obtain accurate prior knowledge, thereby improving the detection accuracy of the detection formula. Furthermore, in the detection formula setting and optimization method provided by this invention, the detection formula strategy and the values ​​of detection parameters are derived through reverse derivation based on defect distribution boundary information and the preset outlier statistical analysis strategy. Therefore, this invention can simultaneously deduce a set of detection parameters (and adjust all parameters at the same time) through reverse derivation, taking into account the coupling relationship between parameters, thus realizing rapid modeling of detection formulas; it avoids repeated parameter adjustments, which can significantly save manpower and time costs; moreover, for the detection of defects in new processes, users do not need to have algorithmic knowledge to determine the strategy of detection formulas and the values ​​of detection parameters of detection formulas.

[0251] It should be noted that the apparatus and methods disclosed in the embodiments herein can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments herein. In this regard, each block in a flowchart or block diagram may represent a module, program, or part of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system to perform the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0252] In addition, the functional modules in the various embodiments of this article can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0253] The above description is merely a description of preferred embodiments of the present invention and is not intended to limit the scope of the invention in any way. Any changes or modifications made by those skilled in the art based on the above disclosure are within the protection scope of the present invention. Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the present invention and its equivalents, the present invention also intends to include these modifications and variations.

Claims

1. A method for setting and optimizing a detection formula, characterized in that, include: A first data sample is labeled to obtain a second data sample. The first data sample includes several detection result data entries. These data entries include basic information and feature data of the detected object. The feature data includes the location information of the detection result on the detected object, and one or more of the following: the process flow information of the detected object, the grayscale information, shape information, and texture information of the detection result. The first data sample is labeled based on the basic information and feature data of the detected object, and it is determined whether the defects indicated by the detection result data are genuine defects or noise data. The second data sample includes the detection result data and a label corresponding to each detection result data entry. Based on the second data sample, data feature distribution information of the detected object is obtained. The data feature distribution information includes feature data information of the detected object and other feature data information besides the feature data information. A preset outlier statistical analysis strategy is used to perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information; and a detection formula is determined based on the preset outlier statistical analysis strategy. Based on the defect distribution boundary information and the preset outlier statistical analysis strategy, the values ​​of the detection parameters of the detection formula are set or optimized through reverse derivation.

2. The method for setting and optimizing the detection formula according to claim 1, characterized in that, The step of labeling the first data sample to obtain the second data sample includes: Obtain the basic information of the detection object corresponding to each detection result data in the first data sample; For each detection result data, based on the basic information of the detection object and the location information of the detection result on the detection object, the original information corresponding to the detection result data on the detection object is obtained; Based on the original information, determine whether the defect indicated by the data information of the detection result is a real defect. If so, mark the detection result data as real defect data; otherwise, mark the detection result data as noise data. The second data sample is obtained based on all the test result data and the label corresponding to each test result data.

3. The method for setting and optimizing the detection formula according to claim 2, characterized in that, The detection object includes a wafer; the basic information of the wafer includes the wafer number, the number of dies it contains, and the basic information of each die; the basic information of the die includes the die number and image information of the die. The step of obtaining the original information corresponding to the detection result data on the detection object based on the basic information of the detection object and the position information of the detection result on the detection object includes: Based on the basic information of the Wafer, obtain the Die number of each Die of the Wafer and the basic information of each Die; Based on the location information of the detection result on the die and the image information of the die, obtain the image information of the detection result corresponding to the detection result data on the die.

4. The method for setting and optimizing the detection formula according to claim 1, characterized in that, The step of obtaining the data feature distribution information of the detected object based on the second data sample includes: A feature data axis and a segmentation data axis are determined, and a feature space is established based on the feature data axis and the segmentation data axis; wherein, the feature data axis represents the feature data information of the detection result data, and the segmentation data axis represents segmentation feature information; wherein, the segmentation feature information includes other feature data information besides that used for the feature data axis; The second data sample is arranged according to the feature space to obtain the data feature distribution information of the detected object.

5. The method for setting and optimizing the detection formula according to claim 4, characterized in that, The feature space includes one or more of the feature data axes and one or more of the segmentation data axes.

6. The method for setting and optimizing the detection formula according to claim 4, characterized in that, The step of arranging the second data samples according to the feature space to obtain the data feature distribution information of the detected object includes: Establish a rectangular coordinate system by taking the feature data axis as the horizontal axis and the segmentation data axis as the vertical axis; In the Cartesian coordinate system, the second data samples are arranged according to the feature values ​​of the feature data information represented by the feature data axis in the horizontal direction and according to the feature values ​​of the feature data information represented by the segmentation data axis in the vertical direction to obtain a defect feature distribution map.

7. The method for setting and optimizing the detection formula according to claim 6, characterized in that, The step of employing a preset outlier statistical analysis strategy to perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information includes: If it is determined whether to automatically search for defect distribution boundary information, then the outlier statistical analysis model is trained according to the selected outlier statistical analysis model to obtain defect distribution boundary information; otherwise, the data segmentation method is used to perform outlier statistical analysis on the data feature distribution information to obtain defect distribution boundary information. The step of training the outlier statistical analysis model includes: training the selected outlier statistical analysis model based on the detection result data and the data feature distribution information until the defect distribution boundary information of the detected object meets the first preset condition. The step of performing outlier statistical analysis on the data feature distribution information using a data segmentation method includes: obtaining at least one first segmentation threshold on the feature data axis and / or the segmentation data axis based on the detection result data and the data feature distribution information; and obtaining the defect distribution boundary information based on the first segmentation threshold until the obtained defect distribution boundary information of the detected object meets a second preset condition.

8. The method for setting and optimizing the detection formula according to claim 7, characterized in that, The segmented data axis represents process flow information; the step of threshold segmenting the feature data axis and / or the segmented data axis based on the detection result data and the data feature distribution information until the defect distribution boundary information of the detected object meets the second preset condition includes: Based on the data feature distribution information and the consistency of the distribution of detection result data labeled as true defect data and labeled as noise data, the first segmentation threshold of the segmented data axis is determined. Based on the data feature distribution information and the consistency of the distribution of detection result data labeled as true defect data and labeled as noise data, a second segmentation threshold for the feature data axis is determined; Based on the first segmentation threshold of the segmentation data axis and the second segmentation threshold of the feature data axis, the defect distribution boundary information of the detected object is obtained.

9. The method for setting and optimizing the detection formula according to claim 7, characterized in that, The aforementioned outlier statistical analysis strategy also includes: an outlier statistical analysis strategy that combines data segmentation and model learning; The outlier statistical analysis strategy combining data segmentation and model learning includes: obtaining at least one first segmentation threshold on the segmentation data axis for the detection result data labeled as true defects based on the data feature distribution information; and training the selected outlier statistical analysis model based on the first segmentation threshold and the data feature distribution information until the defect distribution boundary information of the detected object satisfies a third preset condition.

10. The method for setting and optimizing the detection formula according to claim 1, characterized in that, The step of setting or optimizing the detection parameters of the detection formula by reverse derivation based on the defect distribution boundary information and the preset outlier statistical analysis strategy includes: Based on the preset outlier statistical analysis strategy, determine the reverse derivation strategy; Based on the reverse derivation strategy, determine the input data information of the reverse derivation strategy; Based on the input data information, determine the data distribution model of the detection result data; Based on the data distribution model and the defect distribution boundary information, the detection parameters of the detection formula are determined; Based on the strategy of the detection formula and the input data information derived in reverse, the values ​​of the detection parameters of the detection formula are set or optimized.

11. The method for setting and optimizing the detection formula according to claim 10, characterized in that, The preset outlier statistical analysis strategy is the data segmentation method; According to the data segmentation method, the data distribution density of the detection result data of the detection object is used as the reverse derivation strategy. Based on the reverse derivation strategy of the statistical data distribution density, all detection result data of the detection object are used as the input data information; Based on all the detection results data, it is assumed that the feature values ​​of all the detection results data feature information in the feature space are divided into normal region, noisy region and true defect region; the normal region is the region where the data distribution density is greater than the first density threshold, the noisy region is the region where the data density is less than or equal to the first density threshold and greater than the second density threshold, and the true defect region is the region where the data density is less than or equal to the second density threshold. Based on all detection result data and the labels of all detection result data, calculate the first density threshold and the second density threshold; wherein, the first density threshold is greater than the second density threshold; The displacement parameters of the detection formula are calculated based on the first density threshold, the second density threshold, and the defect distribution boundary information.

12. The method for setting and optimizing the detection formula according to claim 10, characterized in that, The preset outlier statistical analysis strategy is an outlier statistical analysis strategy based on Gaussian model; According to the Gaussian model-based outlier statistical analysis strategy, the Gaussian distribution of the detection result data of the detection object is used as the reverse derivation strategy, and the Gaussian model detection is used as the strategy for the detection formula. Based on the reverse derivation strategy of statistical Gaussian distribution, all detection result data of the detection object and the defect distribution boundary information are used as the input data information. Based on all the detection result data, it is assumed that the feature values ​​of all the feature data information of the detection result data follow a Gaussian distribution in the feature space. The parameters for the Gaussian model detection are determined based on the input data and the defect distribution boundary information.

13. The method for setting and optimizing the detection formula according to claim 10, characterized in that, The preset outlier statistical analysis strategy is a machine learning outlier statistical analysis strategy; Based on the outlier statistical analysis strategy of the machine learning, the density threshold and distance threshold of the detection result data of the detection object are used as the reverse derivation strategy, and the machine learning model is used as the detection formula strategy. Based on the reverse derivation strategy of the density threshold and distance threshold of the detection result data of the detection object, the density and distance of the detection result data of the detection object are used as the input data information. Based on all the detection results data and the defect distribution boundary information, the density parameters and distance parameters of the detection strategy of the machine learning model are derived in reverse.

14. The method for setting and optimizing the detection formula according to any one of claims 1-13, characterized in that, Also includes: Based on the test formula and the values ​​of the test parameters of the test formula, defect analysis is performed on the object to be tested to obtain defect data information of the object to be tested.

15. A device for detecting and optimizing formula settings, characterized in that, include: A true defect and noise labeling unit is configured to label a first data sample to obtain a second data sample. The first data sample includes several detection result data entries, each including basic information and feature data of the detected object. The feature data includes the location information of the detection result on the detected object, and one or more of the following: the process flow information of the detected object, the grayscale information, shape information, and texture information of the detection result. The first data sample is labeled based on the basic information and feature data of the detected object, and it is determined whether the defect indicated by the detection result data is true defect data or noise data. The second data sample includes the detection result data and a label corresponding to each detection result data entry. The feature distribution information acquisition unit is configured to obtain data feature distribution information of the detection object based on the second data sample. The data feature distribution information includes feature data information of the detection object and other feature data information besides the feature data information. The defect distribution boundary acquisition unit is configured to perform outlier statistical analysis on the data feature distribution information using a preset outlier statistical analysis strategy to acquire defect distribution boundary information, and to determine the detection formula according to the preset outlier statistical analysis strategy. The detection parameter setting and optimization unit is configured to set or optimize the values ​​of the detection parameters of the detection formula by reverse derivation based on the defect distribution boundary information and the preset outlier statistical analysis strategy.

16. The detection formula setting and optimization device according to claim 15, characterized in that, Also includes: The test formula application unit is configured to perform defect analysis on the object to be tested based on the test formula and the values ​​of the test parameters of the test formula, and obtain defect data information of the object to be tested.

17. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the detection formula setting and optimization method according to any one of claims 1 to 14.

18. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when executed by a processor, implements the detection formula setting and optimization method according to any one of claims 1 to 14.

Images