A method and device for evaluating measurement uncertainty

Abstract

The embodiment of the present invention provides a method and device for evaluating the uncertainty of measurement, which is applied to the field of measurement analysis, including: obtaining a three-dimensional image of an object to be detected at different angles; the three-dimensional image includes a bright spot structure and a dark detail structure; performing dimensionality reduction processing on the three-dimensional images to be detected at different angles obtained according to a set detection order to obtain a two-dimensional image of the object to be detected; inputting the two-dimensional image into K models to be used respectively, outputting K measurement data and N uncertainties, wherein N is a positive integer not greater than K; determining the target measurement data of the object to be detected and the target uncertainty corresponding to the target measurement data according to the K measurement data and the N uncertainties. The target measurement data and the target uncertainty can be obtained relatively quickly, so as to facilitate the subsequent determination of the accuracy of the target measurement data.

Description

Technical Field

[0001] The present invention relates to the field of measurement analysis, and in particular to a method and device for evaluating measurement uncertainty. Background Art

[0002] X-ray measurement technology uses the interaction between X-rays and matter to obtain information about the internal structure and composition of an object. Since this technology can provide high-resolution information, it has been widely used in medical imaging, industrial inspection, materials science and other fields.

[0003] At present, the MCMC (Markov Chain Monte Carlo) method is generally used to measure the information of the internal structure and composition of an object. Since this technology relies on empirical modeling and complex mathematical calculations, the calculation speed of this technology is slow and the measurement efficiency is low.

[0004] In summary, how to improve the measurement efficiency of X-rays is a technical problem that needs to be solved urgently. Summary of the invention

[0005] The embodiment of the present invention provides a method and device for evaluating measurement uncertainty, so as to solve the problems of slow calculation speed and low measurement efficiency in the prior art.

[0006] In a first aspect, an embodiment of the present invention provides a method for evaluating measurement uncertainty, comprising: obtaining three-dimensional images of an object to be detected at different angles; the three-dimensional images include a bright spot structure and a dark detail structure; performing dimensionality reduction processing on the three-dimensional images to be detected at different angles obtained according to a set detection order to obtain a two-dimensional image of the object to be detected; inputting the two-dimensional images into K models to be used respectively, outputting K measurement data and N uncertainties, wherein N is a positive integer not greater than K; determining target measurement data of the object to be detected and target uncertainty corresponding to the target measurement data based on the K measurement data and the N uncertainties.

[0007] In the above technical solution, by inputting the two-dimensional image into K models to be used, not only the target measurement data but also the target uncertainty corresponding to the target measurement data can be obtained, so as to facilitate the subsequent determination of the accuracy of the target measurement data according to the target uncertainty.

[0008] Optionally, before the two-dimensional images are input into the K models to be used respectively, the following steps are further included:

[0009] According to the parameter adjustment mechanism, the evaluation model is sampled K times to obtain K models to be used. The evaluation model is trained by three-dimensional images of different dimensions of multiple samples. The loss function of the evaluation model includes the loss factor of the measurement data and the loss factor of the uncertainty of the measurement data.

[0010] In the above technical solution, by adjusting the parameter mechanism, the evaluation model is sampled K times to obtain K models to be used. By inputting the two-dimensional image into the K models to be used, the multiple iterative modeling process in the prior art can be replaced, thereby achieving improved calculation speed and improved measurement efficiency.

[0011] Optionally, the three-dimensional images to be detected at different angles are subjected to dimensionality reduction processing to obtain a two-dimensional image of the object to be detected, including: for the three-dimensional image to be detected at any angle, the three-dimensional image to be detected is subjected to dimensionality reduction processing to obtain a two-dimensional graph strip of the angle; M graph strips located in the central area of ​​the two-dimensional graph strip of the angle are used as two-dimensional sub-graphs of the angle; and the two-dimensional sub-graphs of each angle are spliced ​​according to the set detection order to obtain a two-dimensional image.

[0012] In the above technical solution, a two-dimensional image is obtained by performing dimensionality reduction processing on a three-dimensional image, thereby retaining the main structural information of the data, thereby facilitating improving subsequent calculation efficiency.

[0013] Optionally, M strips located in the central area of ​​the two-dimensional strip of the angle are used as the two-dimensional sub-graph of the angle, including: selecting the first (M-1) / 2 strips of the central axis of the two-dimensional image and the last (M-1) / 2 strips of the central axis of the two-dimensional image; and using the selected M strips as the two-dimensional sub-graph.

[0014] Optionally, the evaluation model is trained by three-dimensional images of multiple samples at different angles, including: determining the two-dimensional image of the sample based on the three-dimensional images of the sample at different angles; inputting multiple two-dimensional images of any sample into the initial evaluation model to obtain the measurement value of the sample in each layer and the uncertainty corresponding to the measurement value of each layer; for each layer, determining the layer loss of the layer according to the measurement value of the layer and the label value of the layer; determining the loss factor of the measurement data corresponding to the layer according to the layer loss of the layer and the uncertainty of the layer; determining the loss factor of the uncertainty corresponding to the layer according to the uncertainty of the layer; determining the loss value under the loss function according to the loss factor of the measurement data of each layer and the loss factor of the uncertainty of each layer; adjusting the initial evaluation model according to the loss value to obtain the evaluation model.

[0015] In the above technical solution, the loss value is calculated according to the measurement value of each layer and the uncertainty corresponding to the measurement value of each layer, so as to realize the subsequent adjustment and optimization of the initial evaluation model according to the loss value until the initialized evaluation model meets the training stop requirement. During the training process of the evaluation model, the parameters in the initial evaluation model are adjusted by determining the loss value under the loss function, so that the accuracy of the subsequently adjusted evaluation model is higher.

[0016] Optionally, based on K measurement data and N uncertainties, the target measurement data of the object to be detected and the target uncertainty corresponding to the target measurement data are determined, including: taking the mean or weighted average of the K measurement data as the target measurement data of the object to be detected; determining the target uncertainty corresponding to the target measurement data based on the target measurement data of the object to be detected, N uncertainties, and K measurement data.

[0017] In the above technical solution, the target measurement data can be obtained by averaging or weighted averaging the measurement data output by K models, and the target uncertainty can be determined more accurately through N uncertainties, K measurement data, and target measurement data, thereby achieving not only determining the target measurement data of the object to be detected, but also providing the accuracy of the target measurement data.

[0018] Optionally, the loss function is:

[0019]

[0020] Where x is any sample, y is the label value of the sample, m is the identifier of any layer of the sample, and model(x) m is the measured value of the mth layer, y m is the label value of the mth layer, b m is the uncertainty of the mth level. L(x,y) is the loss value of any sample, and M is the number of sample levels;

[0021] The target uncertainty corresponding to the target measurement data is determined by the following formula:

[0022]

[0023] Where μ is the target measurement data, σ is the target uncertainty, K is the number of models to be used, and b (k) is the uncertainty of the K-th model output, y (k) The measured data output by the Kth model; or

[0024] The loss function is:

[0025]

[0026] The target uncertainty corresponding to the target measurement data is determined by the following formula:

[0027]

[0028] An acquisition unit is used to acquire three-dimensional images of the object to be detected at different angles; the three-dimensional images include bright spot structures and dark detail structures; a processing unit is used to perform dimensionality reduction processing on the three-dimensional images to be detected at different angles acquired according to a set detection sequence to obtain a two-dimensional image of the object to be detected; the two-dimensional images are respectively input into K models to be used, and K measurement data and N uncertainties are output, where N is not greater than K; according to the K measurement data and the N uncertainties, target measurement data of the object to be detected and target uncertainty corresponding to the target measurement data are determined.

[0029] In a second aspect, an embodiment of the present invention provides a device for evaluating measurement uncertainty, comprising: an acquisition unit, used to acquire three-dimensional images of an object to be detected at different angles; the three-dimensional images include bright spot structures and dark detail structures; a processing unit, used to perform dimensionality reduction processing on the three-dimensional images to be detected at different angles acquired according to a set detection order to obtain a two-dimensional image of the object to be detected; the two-dimensional images are respectively input into K models to be used, and K measurement data and N uncertainties are output, where N is not greater than K; based on the K measurement data and N uncertainties, target measurement data of the object to be detected and target uncertainty corresponding to the target measurement data are determined.

[0030] Optionally, the acquisition unit is also used to: sample the evaluation model K times according to the parameter adjustment mechanism to obtain K models to be used, the evaluation model is trained by three-dimensional images of different dimensions of multiple samples, and the loss function of the evaluation model includes a loss factor of the measurement data and a loss factor of the uncertainty of the measurement data.

[0031] Optionally, the processing unit is specifically used to: for the three-dimensional image to be detected at any angle, perform dimensionality reduction processing on the three-dimensional image to be detected at the angle to obtain a two-dimensional graph strip of the angle; use the M graph strips located in the central area of ​​the two-dimensional graph strip of the angle as the two-dimensional sub-graph of the angle; and splice the two-dimensional sub-graphs of each angle according to the set detection order to obtain a two-dimensional image.

[0032] Optionally, the processing unit is specifically used to: select the first (M-1) / 2 strips of the central axis of the two-dimensional image and the last (M-1) / 2 strips of the central axis of the two-dimensional image; and use the selected M strips as the two-dimensional sub-image.

[0033] Optionally, the acquisition unit is specifically used to: determine the two-dimensional image of the sample based on the three-dimensional images of the sample at different angles; input multiple two-dimensional images of any sample into the initial evaluation model to obtain the measurement value of the sample in each layer and the uncertainty corresponding to the measurement value of each layer; for each layer, determine the layer loss of the layer according to the measurement value of the layer and the label value of the layer; determine the loss factor of the measurement data corresponding to the layer according to the layer loss of the layer and the uncertainty of the layer; determine the loss factor of the uncertainty corresponding to the layer according to the uncertainty of the layer; determine the loss value under the loss function according to the loss factor of the measurement data of each layer and the loss factor of the uncertainty of each layer; adjust the initial evaluation model according to the loss value to obtain the evaluation model.

[0034] Optionally, the processing unit is specifically used to: take the mean or weighted average of K measurement data as the target measurement data of the object to be detected; and determine the target uncertainty corresponding to the target measurement data according to the target measurement data of the object to be detected, N uncertainties, and K measurement data.

[0035] Optionally, the loss function is:

[0036]

[0037] Where x is any sample, y is the label value of the sample, m is the identifier of any layer of the sample, and model(x) m is the measured value of the mth layer, y m is the label value of the mth layer, b m is the uncertainty of the mth level. L(x,y) is the loss value of any sample, and M is the number of sample levels;

[0038] The target uncertainty corresponding to the target measurement data is determined by the following formula:

[0039]

[0040] Where μ is the target measurement data, σ is the target uncertainty, K is the number of models to be used, and b (k) is the uncertainty of the K-th model output, y (k) The measured data output by the Kth model; or

[0041] The loss function is:

[0042]

[0043] The target uncertainty corresponding to the target measurement data is determined by the following formula:

[0044]

[0045] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising at least one processor and at least one memory, wherein the memory stores a computer program, and when the program is executed by the processor, the processor executes a method for evaluating measurement uncertainty according to the first aspect above.

[0046] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program executable by a computer device. When the program runs on the computer device, the computer device executes a method for evaluating measurement uncertainty in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A flow chart of a method for evaluating measurement uncertainty provided by an embodiment of the present invention;

[0048] Figure 2 A flow chart of a method for training an evaluation model provided by an embodiment of the present invention;

[0049] Figure 3 A flow chart of a method for obtaining a two-dimensional image of an object to be detected provided by an embodiment of the present invention;

[0050] Figure 4 A schematic diagram of a two-dimensional image provided by an embodiment of the present invention;

[0051] Figure 5 A flow chart of a method for determining target measurement data and target uncertainty corresponding to the target measurement data provided by an embodiment of the present invention;

[0052] Figure 6 A schematic diagram of the structure of a device for evaluating measurement uncertainty provided by an embodiment of the present invention;

[0053] Figure 7 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Integrated circuit (IC) nano-measurement technology is an effective technical means to ensure the quality and consistency of IC manufacturing. In the process of mass manufacturing, it is necessary to perform fast, non-destructive and accurate measurement of the measurement data of IC nano-structures. The measurement data includes not only measurement data such as line width, line height, and side wall angle, but also important features such as line width roughness (LER) and line edge roughness (LWR). In recent years, in order to ensure that the IC structure still maintains excellent performance while being continuously reduced, its structural design has gradually shifted from simple planar structures to complex three-dimensional structures. Among them, the critical dimensions of complex three-dimensional structures are smaller, and more measurement data need to be measured, which puts higher requirements on IC nano-measurement technology.

[0056] An embodiment of the present invention provides a method for evaluating measurement uncertainty, which is used to measure measurement data of an object to be detected, wherein the object to be detected may be a field effect transistor of an integrated circuit, or may be other nanostructured devices, which is not limited herein.

[0057] like Figure 1 FIG. 1 is a flow chart of a method for evaluating measurement uncertainty provided by an embodiment of the present invention, the method comprising the following steps:

[0058] Step 101 : acquiring three-dimensional images of the object to be detected at different angles.

[0059] In an embodiment of the present invention, three-dimensional images of the object to be detected at different angles are obtained, so that the structural information of the object to be detected can be more accurately determined based on the three-dimensional images to be detected at multiple angles. The three-dimensional image has multiple bright spot structures and dark detail structures. For example, if the object to be detected is a field effect transistor of an integrated circuit, then the three-dimensional image is a three-dimensional image of the field effect transistor of the integrated circuit taken at any angle, wherein the three-dimensional image has multiple bright spot structures and dark detail structures, and the structural information and shape information of the field effect transistor of the integrated circuit can be represented by the above structure.

[0060] Step 102 , performing dimensionality reduction processing on the three-dimensional images to be detected at different angles acquired according to the set detection sequence to obtain a two-dimensional image of the object to be detected.

[0061] In the embodiment of the present invention, since the size of a multidimensional image is larger than that of a two-dimensional image and it usually has a higher resolution, it will take a long time to load and analyze the multidimensional image. Therefore, by performing dimensionality reduction processing on the three-dimensional images to be detected at different angles obtained in a set detection order, a two-dimensional image is obtained, so that the main structural information and shape information of the data can be retained, which is convenient for improving the subsequent calculation efficiency.

[0062] Step 103 , input the two-dimensional images into K models to be used respectively, and output K measurement data and N uncertainties.

[0063] In an embodiment of the present invention, the two-dimensional image is respectively input into K models to be used, and K measurement data and K uncertainties are obtained, wherein N is a positive integer not greater than K. According to the K measurement data and N uncertainties, it is convenient to subsequently determine the target measurement data and the target uncertainty corresponding to the target measurement data, so as to achieve that the accuracy of the target measurement data can be determined while determining the target measurement data. For example, if the two-dimensional image is respectively input into 5 models to be used, and the uncertainties output by two of the models to be used are the same, then 5 measurement data and 4 uncertainties are obtained. For another example, if the two-dimensional image is respectively input into 5 models to be used, and the uncertainties output by the 5 models to be used are all different, then 5 measurement data and 5 uncertainties are obtained.

[0064] Step 104 : determining target measurement data of the object to be detected and target uncertainty corresponding to the target measurement data according to the K measurement data and the N uncertainties.

[0065] In the embodiment of the present invention, the target measurement data of the object to be detected can be determined more accurately by combining K measurement data, and the target uncertainty corresponding to the target measurement data can be determined according to N uncertainties, so that not only the target measurement data of the object to be detected can be determined, but also the accuracy of the target measurement data of the object to be detected can be determined, which is convenient for subsequent optimization and adjustment of the evaluation model according to the accuracy of the target measurement data, thereby improving the measurement accuracy of the evaluation model.

[0066] It can be seen from the above steps 101 to 104 that by inputting the two-dimensional image into the K models to be used, not only the target measurement data but also the target uncertainty corresponding to the target measurement data can be obtained, thereby facilitating the subsequent determination of the accuracy of the target measurement data according to the target uncertainty.

[0067] In the embodiment of the present invention, in order to determine the target measurement data and the target uncertainty corresponding to the target measurement data, first, K models to be used are determined, wherein the K models to be used are obtained by sampling the evaluation model K times according to the adjustment parameter mechanism of the evaluation model. Therefore, the following describes how to train the obtained evaluation model.

[0068] like Figure 2 FIG. 1 is a flow chart of a method for training an evaluation model provided by an embodiment of the present invention, and the method comprises the following steps:

[0069] Step 201 : for each sample, determine a two-dimensional image of the sample according to three-dimensional images of the sample at different angles.

[0070] In the embodiment of the present invention, for each sample, dimensionality reduction processing is performed based on the three-dimensional images of the sample at different angles to obtain a two-dimensional image of the sample, thereby retaining the main structural information of the data and facilitating the improvement of subsequent calculation efficiency.

[0071] Step 202: Input a two-dimensional image of any sample into the initial evaluation model to obtain the measurement value of the sample at each layer and the uncertainty corresponding to the measurement value of each layer.

[0072] In an embodiment of the present invention, a two-dimensional image of any sample is input into the initial evaluation model to obtain the measurement value of the sample in each layer and the uncertainty corresponding to the measurement value of each layer, wherein the initial evaluation model can be at least one of FCN, Unet, resnet, dilated convolution, Enet, densenet, and R-CNN, and at least one convolution layer in the network model is packaged by at least one of dropout, concrete dropout, and guassian dropout, and an FC layer and a softplus layer that meet the output length are added to the last layer of the network. The output of the FC layer is the measurement value, and the output of the softplus layer is the uncertainty. For example, if the surface of the sample is uneven, in order to obtain more accurate measurement data, the initial evaluation model will stratify the sample, input the measurement value of the sample in each layer and the uncertainty corresponding to the measurement value of each layer.

[0073] Step 203: Determine the loss value under the loss function according to the measurement value of each layer and the uncertainty corresponding to the measurement value of each layer.

[0074] In an embodiment of the present invention, in order to improve the accuracy of the measurement of the initial evaluation model, it is necessary to determine the loss value under the loss function according to the measurement value of each layer and the uncertainty corresponding to the measurement value of each layer, so as to facilitate the subsequent adjustment of the initial evaluation model according to the loss value under the loss function. For each layer, the layer loss of the layer is determined according to the measurement value of the layer and the label value of the layer; the loss factor of the measurement data corresponding to the layer is determined according to the layer loss of the layer and the uncertainty of the layer; the loss factor of the uncertainty corresponding to the layer is determined according to the uncertainty of the layer; the loss value under the loss function is determined according to the loss factor of the measurement data of each layer and the loss factor of the uncertainty of each layer. There are many methods for determining the loss value under the loss function, which can be to determine the loss value according to the Gaussian distribution likelihood function, or to determine the loss value according to the Laplace likelihood function, or other methods, which are not limited here. The following introduces how to determine the loss value according to the Gaussian distribution likelihood function.

[0075] The formula for determining the loss value based on the Gaussian distribution likelihood function is as follows, see Formula 1:

[0076]

[0077] Where x is any sample, y is the label value of the sample, m is the identifier of any layer of the sample, and model(x) m is the measured value of the mth layer, y m is the label value of the mth layer, b m is the uncertainty of the mth layer. L(x,y) is the loss value of any sample, M is the number of sample levels, is the loss factor of the measured data corresponding to the mth layer, is the loss factor of uncertainty corresponding to the mth layer.

[0078] The following describes how to determine the loss value based on the Laplace likelihood function.

[0079] The formula for determining the loss value based on the Laplace likelihood function is as follows, see Formula 2:

[0080]

[0081] Where x is any sample, y is the label value of the sample, m is the identifier of any layer of the sample, and model(x) m is the measured value of the mth layer, y m is the label value of the mth layer, b m is the uncertainty of the mth layer. L(x,y) is the loss value of any sample, M is the number of sample levels, is the loss factor of the measured data corresponding to the mth layer, is the loss factor of uncertainty corresponding to the mth layer.

[0082] Step 204, adjust the initial evaluation model according to the loss value until the training stop requirement is met to obtain the evaluation model.

[0083] In the embodiment of the present invention, the loss function is suitable for evaluating the difference between the measured value of each layer of the evaluation model and the label value of each layer, wherein the smaller the loss value, the closer the measured value of each layer output by the evaluation model is to the label value of the corresponding layer, and the higher the accuracy of the evaluation model. Therefore, according to the loss value, the initial evaluation model is adjusted until the training stop requirement is met to obtain the evaluation model.

[0084] It can be seen from the above steps 201 to 204 that by training the initialization evaluation model, the accuracy of the trained evaluation model is improved, so that target measurement data can be obtained more accurately in the subsequent process.

[0085] In the embodiment of the present invention, after obtaining the evaluation model, the evaluation model samples the evaluation model K times according to the parameter adjustment mechanism to obtain K models to be used. Specifically, the dropout, concrete dropout or guassian dropout module is not closed during the sampling process, and the sampling is performed K times to obtain the uncertainty evaluation model sequence 5. By adjusting the parameter mechanism, the evaluation model is sampled K times to obtain K models to be used. By inputting the two-dimensional image into the K models to be used, the multiple iterative modeling process in the prior art can be replaced, thereby achieving improved calculation speed and improved measurement efficiency.

[0086] Optionally, after obtaining the three-dimensional images of the object to be detected at different angles, it is necessary to first perform dimensionality reduction processing on the three-dimensional images to be detected to obtain two-dimensional images, and then input the two-dimensional images into K models to be used, so as to obtain target measurement data and the uncertainty corresponding to the target measurement data. The following describes how to perform dimensionality reduction processing on the three-dimensional images to be detected to obtain two-dimensional images.

[0087] like Figure 3 FIG. 1 is a flow chart of a method for obtaining a two-dimensional image of an object to be detected provided by an embodiment of the present invention. The method comprises the following steps:

[0088] Step 301, for any angle of the to-be-detected 3D image, perform dimensionality reduction processing on the to-be-detected 3D image of the angle to obtain a 2D graph strip of the angle, and use M graph strips in the center area of ​​the 2D graph strip of the angle as the 2D subgraph of the angle.

[0089] In an embodiment of the present invention, for a three-dimensional image to be detected at any angle, a dimensionality reduction process is performed on the three-dimensional image to be detected to obtain a two-dimensional bar of the angle, and then M bars located in the center area of ​​the two-dimensional bar of the angle are used as a two-dimensional sub-graph of the angle. For example, if there are a total of 100 three-dimensional images to be detected at different angles, M is 5, and the three-dimensional image to be detected at each angle is subjected to dimensionality reduction process to obtain 250 two-dimensional bars, wherein 5 two-dimensional bars are taken for each angle, and the principle of selecting the two-dimensional bars is based on the central axis of the two-dimensional bar, 2 two-dimensional bars are taken forward, and 2 two-dimensional bars are taken backward, and the 5 two-dimensional bars are the two-dimensional sub-graph of the angle.

[0090] Step 302: splice the two-dimensional sub-images at each angle according to the set detection order to obtain a two-dimensional image.

[0091] In the embodiment of the present invention, according to step 301, the two-dimensional sub-images at each angle can be determined, and then the two-dimensional sub-images at each angle are spliced ​​according to the set detection order to obtain a two-dimensional image. Figure 4 .

[0092] It can be seen from the above steps 301 to 302 that by performing dimensionality reduction processing on a three-dimensional image to obtain a two-dimensional image, the main structural information of the data can be retained, which facilitates improving the subsequent calculation efficiency.

[0093] like Figure 5 FIG. 1 is a flow chart of a method for determining target measurement data and target uncertainty corresponding to the target measurement data provided by an embodiment of the present invention. The method comprises the following steps:

[0094] Step 501: taking the mean or weighted average of K measurement data as target measurement data of the object to be detected.

[0095] In the embodiment of the present invention, the average of K measurement data is used as the target measurement data of the object to be detected.

[0096]

[0097] Among them, μ is the target measurement data, K is the number of models to be used, and y (k) The measurement data output by the Kth model.

[0098] Optionally, a weighted average of K measurement data may be used as target measurement data of the object to be detected.

[0099] Step 502: Determine the target uncertainty corresponding to the target measurement data according to the target measurement data of the object to be detected, N uncertainties, and K measurement data.

[0100] In the embodiment of the present invention, it should be noted that if in the above step 203, the loss value is determined according to the Gaussian distribution likelihood function, the formula for determining the target uncertainty is shown in Formula 3:

[0101]

[0102] Where μ is the target measurement data, σ is the target uncertainty, K is the number of models to be used, and b (k) is the uncertainty of the K-th model output, y (k) The measured data (y (k) -μ) 2 .

[0103] If in the above step 203, the loss value is determined according to the Laplace likelihood function, the formula for determining the target uncertainty is shown in Formula 4:

[0104]

[0105] Where μ is the target measurement data, σ is the target uncertainty, K is the number of models to be used, and b (k)is the uncertainty of the K-th model output, y (k) The measured data (y (k) -μ) 2 .

[0106] It can be seen from the above steps 501 to 502 that by taking the mean or weighted average of the measurement data output by the K models as the target measurement data, and based on the target measurement data of the object to be detected, N uncertainties, and K measurement data, the target uncertainty can be determined more accurately, thereby achieving not only determining the target measurement data of the object to be detected, but also providing the accuracy of the target measurement data.

[0107] Based on the same technical concept, an embodiment of the present invention provides a device for evaluating measurement uncertainty, such as Figure 6 As shown, the device 600 includes: an acquisition unit 601, which is used to acquire a three-dimensional image of the object to be detected at different angles; the three-dimensional image includes a bright spot structure and a dark detail structure; a processing unit 602, which is used to perform dimensionality reduction processing on the three-dimensional images to be detected at different angles acquired according to a set detection order to obtain a two-dimensional image of the object to be detected; the two-dimensional images are respectively input into K models to be used, and K measurement data and N uncertainties are output, where N is not greater than K; according to the K measurement data and the N uncertainties, the target measurement data of the object to be detected and the target uncertainty corresponding to the target measurement data are determined.

[0108] Optionally, the acquisition unit 601 is also used to: sample the evaluation model K times according to the parameter adjustment mechanism to obtain K models to be used, the evaluation model is trained by three-dimensional images of different dimensions of multiple samples, and the loss function of the evaluation model includes a loss factor of the measurement data and a loss factor of the uncertainty of the measurement data.

[0109] Optionally, the processing unit 602 is specifically used to: for the three-dimensional image to be detected at any angle, perform dimensionality reduction processing on the three-dimensional image to be detected at the angle to obtain a two-dimensional graph strip of the angle; use the M graph strips located in the central area of ​​the two-dimensional graph strip of the angle as the two-dimensional sub-graph of the angle; and splice the two-dimensional sub-graphs of each angle according to the set detection order to obtain a two-dimensional image.

[0110] Optionally, the processing unit 602 is specifically used to: select the first (M-1) / 2 strips of the central axis of the two-dimensional image and the last (M-1) / 2 strips of the central axis of the two-dimensional image; and use the selected M strips as the two-dimensional sub-image.

[0111] Optionally, the acquisition unit 601 is specifically used to: determine the two-dimensional image of the sample based on the three-dimensional images of the sample at different angles; input multiple two-dimensional images of any sample into the initial evaluation model to obtain the measurement value of the sample in each layer and the uncertainty corresponding to the measurement value of each layer; for each layer, determine the layer loss of the layer according to the measurement value of the layer and the label value of the layer; determine the loss factor of the measurement data corresponding to the layer according to the layer loss of the layer and the uncertainty of the layer; determine the loss factor of the uncertainty corresponding to the layer according to the uncertainty of the layer; determine the loss value under the loss function according to the loss factor of the measurement data of each layer and the loss factor of the uncertainty of each layer; adjust the initial evaluation model according to the loss value to obtain the evaluation model.

[0112] Optionally, the processing unit 602 is specifically used to: use the mean or weighted average of the K measurement data as the target measurement data of the object to be detected; and determine the target uncertainty corresponding to the target measurement data according to the target measurement data of the object to be detected, N uncertainties, and K measurement data.

[0113] Optionally, the loss function is:

[0114]

[0115] Where x is any sample, y is the label value of the sample, m is the identifier of any layer of the sample, and model(x) m is the measured value of the mth layer, y m is the label value of the mth layer, b m is the uncertainty of the mth level. L(x,y) is the loss value of any sample, and M is the number of sample levels;

[0116] The target uncertainty corresponding to the target measurement data is determined by the following formula:

[0117]

[0118] Where μ is the target measurement data, σ is the target uncertainty, K is the number of models to be used, and b (k) is the uncertainty of the K-th model output, y (k) The measured data output by the Kth model; or

[0119] The loss function is:

[0120]

[0121] The target uncertainty corresponding to the target measurement data is determined by the following formula:

[0122]

[0123]

[0124]

[0125] Based on the same technical concept, the embodiment of the present application also provides an electronic device, such as Figure 7 As shown, the electronic device 700 includes at least one processor 701 and a memory 702 connected to the at least one processor. The specific connection medium between the processor 701 and the memory 702 is not limited in the embodiment of the present application. Figure 7 For example, the processor 701 and the memory 702 are connected via a bus. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0126] In the embodiment of the present application, the memory 702 stores instructions that can be executed by at least one processor 701. The at least one processor 701 can execute the steps included in the aforementioned method for evaluating measurement uncertainty by executing the instructions stored in the memory 702.

[0127] The processor 701 is the control center of the computing device, and can use various interfaces and lines to connect various parts of the computing device, and realize data processing by running or executing instructions stored in the memory 702 and calling data stored in the memory 702. The processor 701 can be a general-purpose processor, such as a central processing unit (CPU), a digital signal processor, an application-specific integrated circuit (ASIC), a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, and can realize or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application.

[0128] The memory 702 is a non-volatile computer-readable storage medium that can be used to store non-volatile software programs, non-volatile computer executable programs, and modules. The memory 702 may include at least one type of storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory, a random access memory (RAM), a static random access memory (SRAM), a programmable read-only memory (PROM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic memory, a disk, an optical disk, and the like.

[0129] Based on the same technical concept, an embodiment of the present application also provides a computer-readable storage medium, which stores a computer program that can be executed by a computing device. When the program runs on an electronic device, the electronic device executes the steps of the above-mentioned method for evaluating measurement uncertainty.

[0130] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0131] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0132] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0133] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0134] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0135] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for evaluating the uncertainty of measurement, characterized in that include: Acquire a three-dimensional image of the object to be detected at different angles; the three-dimensional image includes a bright spot structure and a dark detail structure; Performing dimensionality reduction processing on the three-dimensional images to be detected at different angles obtained in the detection order to obtain two-dimensional sub-images of the angles, and splicing the two-dimensional sub-images of each angle to obtain a two-dimensional image of the object to be detected; Input the two-dimensional images into the K models to be used respectively, and output K measurement data and N uncertainties, where N is a positive integer less than or equal to K; Determining target measurement data of the object to be detected and target uncertainty corresponding to the target measurement data according to the K measurement data and the N uncertainties; Before the two-dimensional images are respectively input into the K models to be used, the method further includes: According to the parameter adjustment mechanism, the evaluation model is sampled K times to obtain the K models to be used. The evaluation model is trained by three-dimensional images of different dimensions of multiple samples. The loss function of the evaluation model includes a loss factor of the measurement data and a loss factor of the uncertainty of the measurement data.

2. The method according to claim 1, characterized in that The three-dimensional images to be detected at different angles obtained in the detection order are subjected to dimensionality reduction processing to obtain two-dimensional sub-images of the angles, and the two-dimensional sub-images of the angles are spliced ​​to obtain a two-dimensional image of the object to be detected, including: For a three-dimensional image to be detected at any angle, a dimensionality reduction process is performed on the three-dimensional image to be detected at the angle to obtain a two-dimensional image strip at the angle; Using M graph strips located in the central area of ​​the two-dimensional graph strip of the angle as the two-dimensional subgraph of the angle; According to the detection order, the two-dimensional sub-images at each angle are spliced ​​to obtain the two-dimensional image.

3. The method according to claim 2, characterized in that The method of using M graph strips located in a central area of ​​the two-dimensional graph strip of the angle as the two-dimensional subgraph of the angle includes: Select the first (M-1) / 2 strips of the central axis of the two-dimensional image and the last (M-1) / 2 strips of the central axis of the two-dimensional image; and use the selected M strips as the two-dimensional sub-image.

4. The method according to claim 1, characterized in that The evaluation model is trained by three-dimensional images of multiple samples at different angles, including: Determining a two-dimensional image of the sample according to three-dimensional images of the sample at different angles; Input multiple two-dimensional images of any sample into the initial evaluation model to obtain the measurement value of the sample in each layer and the uncertainty corresponding to the measurement value of each layer; For each layer, determining the layer loss of the layer according to the measurement value of the layer and the label value of the layer; Determine the loss factor of the measurement data corresponding to the layer according to the layer loss of the layer and the uncertainty of the layer; determine the loss factor of the uncertainty corresponding to the layer according to the uncertainty of the layer; Determining the loss value under the loss function according to the loss factor of the measurement data of each layer and the loss factor of the uncertainty of each layer; The initial evaluation model is adjusted according to the loss value to obtain the evaluation model.

5. The method according to any one of claims 1 to 4, characterized in that: Determining target measurement data of the object to be detected and target uncertainty corresponding to the target measurement data according to the K measurement data and the N uncertainties includes: Taking the mean or weighted average of the K measurement data as the target measurement data of the object to be detected; The target uncertainty corresponding to the target measurement data is determined according to the target measurement data of the object to be detected, the N uncertainties, and the K measurement data.

6. The method according to claim 5, characterized in that The loss function is: Where x is any sample, y is the label value of the sample, m is the identifier of any layer of the sample, and model(x) m is the measured value of the mth layer, y m is the label value of the mth layer, b m is the uncertainty of the mth layer; L(x,y) is the loss value of any sample, and M is the number of sample levels; The target uncertainty corresponding to the target measurement data is determined by the following formula: Wherein, μ is the target measurement data, σ is the target uncertainty, K is the number of models to be used, and b (k) is the uncertainty of the K-th model output, y (k) The measured data output by the Kth model; or The loss function is: The target uncertainty corresponding to the target measurement data is determined by the following formula:

7. A device for evaluating measurement uncertainty, characterized in that include: An acquisition unit, used for acquiring a three-dimensional image of the object to be detected at different angles; the three-dimensional image includes a bright spot structure and a dark detail structure; A processing unit is used to perform dimensionality reduction processing on the three-dimensional images to be detected at different angles obtained according to a set detection order to obtain two-dimensional sub-images of the angles, and to splice the two-dimensional sub-images of each angle to obtain a two-dimensional image of the object to be detected; input the two-dimensional images into the K models to be used respectively, and output K measurement data and N uncertainties, where N is not greater than K; determine the target measurement data of the object to be detected and the target uncertainty corresponding to the target measurement data according to the K measurement data and the N uncertainties; The processing unit is also used to: sample the evaluation model K times according to the parameter adjustment mechanism to obtain the K models to be used, the evaluation model is trained by three-dimensional images of different dimensions of multiple samples, and the loss function of the evaluation model includes a loss factor of the measurement data and a loss factor of the uncertainty of the measurement data.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images