A method for obtaining an optical parameter of a detection object and a related device

Abstract

Embodiments of the present application provide a method for obtaining optical parameters of a detection object, which can automatically generate the optical parameters of the detection object based on the ncs spectral image of the detection object, thereby improving the convenience from the ncs spectral image to the output of the optical parameters. The method comprises: obtaining the ncs spectral image of the detection object, the ncs spectral image representing the projection images of the spectral image of the detection object in three mutually orthogonal directions respectively; obtaining a plurality of sets of optical parameter values corresponding to the ncs spectral image of the detection object, wherein each set of optical parameters comprises at least two types of optical parameters; inputting the plurality of sets of optical parameter values into a first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model; determining, among the plurality of predicted ncs spectral images, a target ncs spectral image with the minimum error from the ncs spectral image of the detection object; and regarding the target set of optical parameters corresponding to the target ncs spectral image as the optical parameters of the detection object.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a method for obtaining optical parameters of a detection object and related devices. BACKGROUND

[0002] When detecting the detection object, the detection device can use the detection device to perform optical imaging on the surface of the detection object, and the computer can perform image processing on the detected optical image and extract the detection image. These image-processed image data can be stored in the computer. The above process is implemented in the detection device, and the spectral image of the detection object can be quickly displayed in the computer.

[0003] Generally, the spectral image of each detection object corresponds to a set of optical parameters of the detection object, such as the thickness, top width, bottom width, optical path angle, and azimuth angle of the detection object. However, in the prior art, after obtaining the spectral image of each detection object, the output of the spectral image of the detection object to the optical parameters cannot be automatically realized. SUMMARY

[0004] The embodiment of the present application provides a method for obtaining optical parameters of a detection object, which can automatically generate the optical parameters of the detection object based on the ncs spectral image of the detection object, thereby improving the convenience of outputting from the ncs spectral image to the optical parameters.

[0005] The first aspect of the embodiment of the present application provides a method for obtaining optical parameters of a detection object, comprising:

[0006] Obtaining the ncs spectral image of the detection object, wherein the ncs spectral image represents the projection images of the spectral image of the detection object in three mutually orthogonal directions, respectively;

[0007] Obtaining a plurality of sets of optical parameter values corresponding to the ncs spectral image of the detection object, wherein each set of optical parameters includes at least two types of optical parameters;

[0008] Inputting the plurality of sets of optical parameter values into a first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model;

[0009] In the plurality of predicted ncs spectral images, determining a target ncs spectral image with the minimum error with the ncs spectral image of the detection object;

[0010] Regarding the target optical parameter set corresponding to the target ncs spectral image as the optical parameters of the detection object.

[0011] Preferably, in the plurality of predicted ncs spectral images, a target ncs spectral image with minimum error to the ncs spectral image of the detection object is determined, comprising:

[0012] The plurality of predicted ncs spectral images are subjected to regression analysis with the ncs spectral image of the detection object to obtain a target ncs spectral image with minimum error to the ncs spectral image of the detection object in the plurality of predicted ncs spectral images.

[0013] Preferably, in the method, the plurality of sets of optical parameter values are input into the first neural network model, and the method further comprises:

[0014] A first training sample is obtained, and the first training sample comprises a plurality of ncs spectral images of a plurality of detection objects and a plurality of sets of optical parameters corresponding to the plurality of ncs spectral images.

[0015] The plurality of ncs spectral images of the plurality of detection objects and the plurality of sets of optical parameters corresponding to the plurality of ncs spectral images are input into the first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model.

[0016] According to a preset loss function, a loss between the plurality of ncs spectral images of the plurality of detection objects and the plurality of predicted ncs spectral images is calculated.

[0017] According to the loss and a back propagation algorithm, the first neural network model is trained until the loss reaches a target loss.

[0018] Preferably, the first neural network model comprises a convolution module, a normalization layer, a dropout layer, and a full connection layer.

[0019] The plurality of sets of optical parameter values are input into the first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model, comprising:

[0020] The plurality of sets of optical parameter values are input into the convolution module to extract data features of the plurality of sets of optical parameter values.

[0021] The data features are input into the normalization layer to normalize the data features.

[0022] The normalized data features are input into the dropout layer to de-sparse the normalized data features.

[0023] The de-sparse data features are input into the full connection layer to fit the de-sparse data features to obtain the plurality of predicted ncs spectral images.

[0024] The second aspect of the embodiments of the present application provides a method for obtaining optical parameters of a detection object, comprising:

[0025] obtaining an ncs spectral image of the detection object, wherein the ncs spectral image comprises projection images of a spectral image of the detection object in three mutually orthogonal directions;

[0026] inputting the ncs spectral image into a second neural network model to obtain a group of optical parameters of the detection object output by the second neural network model, wherein the second neural network model is used to convert the ncs spectral image of the detection object into a group of optical parameters of the detection object.

[0027] Preferably, before the ncs spectral image is input into the second neural network model, the method further comprises:

[0028] obtaining second training sample data, wherein the second training sample data comprises an ncs spectral image of each of a plurality of detection objects and a group of optical parameters corresponding to the ncs spectral image;

[0029] inputting the second training sample data into the second neural network model for training until the output of the second neural network model meets a target loss.

[0030] Preferably, the inputting the second training sample data into the second neural network model for training comprises:

[0031] inputting a plurality of ncs spectral images of a plurality of detection objects and a plurality of groups of optical parameters corresponding to the plurality of ncs spectral images into the second neural network model to obtain a plurality of groups of predicted optical parameters output by the second neural network model;

[0032] calculating a loss between the plurality of groups of optical parameters of the plurality of detection objects and the plurality of groups of predicted optical parameters according to a preset loss function;

[0033] training the second neural network model according to the loss and a back propagation algorithm until the loss of the second neural network model meets a target loss.

[0034] Preferably, the inputting the second training sample data into the second neural network model for training comprises:

[0035] deploying one second neural network model on each of n working nodes, and the initialization parameters of the second neural network models are the same, wherein n is greater than or equal to 2;

[0036] dividing the training sample data into n groups of sample data evenly.

[0037] updating parameters of the second neural network model by using a set of sample data on each worker node;

[0038] obtaining n updated second neural network models after updating the parameters of the second neural network model m times on each worker node, 5≤m≤15;

[0039] calculating the average of the model parameters of the n updated second neural network models;

[0040] taking the average of the model parameters as the final model parameters of the second neural network model.

[0041] Preferably, the second neural network model comprises a convolution module, a dropout layer, a full connection layer and a normalization layer.

[0042] inputting the ncs spectral image into the second neural network model to obtain a set of optical parameters of the detection object output by the second neural network model, comprising:

[0043] inputting the ncs spectral image into the convolution module to extract image features of the ncs spectral image;

[0044] inputting the image features into the dropout layer to perform desparsing on the image features;

[0045] inputting the desparsed image features into the full connection layer to fit the desparsed image features;

[0046] inputting the fitted image features into the normalization layer to obtain a set of optical parameters of the detection object output by the second neural network model.

[0047] The third aspect of the embodiment of the present application provides a device for obtaining optical parameters of a detection object, comprising:

[0048] a first obtaining unit configured to obtain an ncs spectral image of the detection object, wherein the ncs spectral image represents projection images of a spectral image of the detection object in three mutually orthogonal directions;

[0049] the first obtaining unit is further configured to obtain a plurality of sets of optical parameter values corresponding to the ncs spectral image of the detection object, wherein each set of optical parameters comprises at least two types of optical parameters;

[0050] a first input-output unit configured to input the plurality of sets of optical parameter values into a first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model;

[0051] The first determining unit is configured to determine a target ncs spectral image with the minimum error from the ncs spectral image of the detection object from the plurality of predicted ncs spectral images.

[0052] The first determining unit is further configured to regard a target optical parameter group corresponding to the target ncs spectral image as the optical parameter group of the detection object.

[0053] Preferably, the first determining unit is specifically configured to:

[0054] perform regression analysis on the plurality of predicted ncs spectral images and the ncs spectral image of the detection object to obtain the target ncs spectral image with the minimum error from the ncs spectral image of the detection object from the plurality of predicted ncs spectral images.

[0055] Preferably, the first obtaining unit is further configured to:

[0056] obtain a first training sample, wherein the first training sample comprises a plurality of ncs spectral images of a plurality of detection objects and a plurality of optical parameter groups corresponding to the plurality of ncs spectral images.

[0057] The first input and output unit is further configured to input the plurality of ncs spectral images of the plurality of detection objects and the plurality of optical parameter groups corresponding to the plurality of ncs spectral images to the first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model.

[0058] Preferably, the device further comprises:

[0059] The computing unit is configured to calculate a loss between the plurality of ncs spectral images of the plurality of detection objects and the plurality of predicted ncs spectral images according to a preset loss function.

[0060] The training unit is configured to train the first neural network model according to the loss and a back propagation algorithm until the loss reaches a target loss.

[0061] Preferably, the first neural network model comprises a convolution module, a normalization layer, a dropout layer and a full connection layer.

[0062] The first input and output unit is specifically configured to:

[0063] input the plurality of optical parameter groups to the convolution module to extract data features of the plurality of optical parameter groups.

[0064] input the data features to the normalization layer to normalize the data features.

[0065] inputting the normalized data features into the dropout layer to de-sparsify the normalized data features;

[0066] inputting the de-sparsified data features into the fully connected layer to fit the de-sparsified data features to obtain the plurality of predicted ncs spectral images.

[0067] The fourth aspect of the embodiment of the present application provides a device for obtaining optical parameters of a detection object, comprising:

[0068] a second obtaining unit configured to obtain an ncs spectral image of the detection object, the ncs spectral image comprising projection images of a spectral image of the detection object in three mutually orthogonal directions;

[0069] a second input-output unit configured to input the ncs spectral image into a second neural network model to obtain a group of optical parameters of the detection object output by the second neural network model, wherein the second neural network model is configured to convert the ncs spectral image of the detection object into the group of optical parameters of the detection object.

[0070] Preferably, the second obtaining unit is further configured to:

[0071] obtain second training sample data, the second training sample data comprising an ncs spectral image of each of a plurality of detection objects and a group of optical parameters corresponding to the ncs spectral image;

[0072] The device further comprises:

[0073] a training unit configured to input the second training sample data into the second neural network model for training until the output of the second neural network model satisfies a target loss.

[0074] Preferably, the training unit is specifically configured to:

[0075] input a plurality of ncs spectral images of a plurality of detection objects and a plurality of groups of optical parameters corresponding to the plurality of ncs spectral images into the second neural network model to obtain a plurality of groups of predicted optical parameters output by the second neural network model;

[0076] calculate a loss between the plurality of groups of optical parameters of the plurality of detection objects and the plurality of groups of predicted optical parameters according to a preset loss function;

[0077] train the second neural network model according to the loss and a back propagation algorithm until the loss of the second neural network model satisfies a target loss.

[0078] Preferably, the training unit is specifically configured to:

[0079] deploying one of the second neural network models on each of the n working nodes, and initialization parameters of the second neural network models are the same, wherein n is greater than or equal to 2;

[0080] dividing the training sample data into n groups of sample data evenly;

[0081] updating parameters of the second neural network model on each working node by using one group of sample data;

[0082] after updating the parameters of the second neural network model m times on each working node, obtaining n updated second neural network models, 5≤m≤15;

[0083] calculating an average of model parameters of the n updated second neural network models;

[0084] taking the average of the model parameters as final model parameters of the second neural network model.

[0085] Preferably, the second neural network model comprises a convolution module, a dropout layer, a full connection layer and a normalization layer.

[0086] a second input and output unit, specifically configured to:

[0087] inputting the ncs spectral image into the convolution module to extract image features of the ncs spectral image;

[0088] inputting the image features into the dropout layer to perform de-sparsification on the image features;

[0089] inputting the de-sparsified image features into the full connection layer to fit the de-sparsified image features;

[0090] inputting the fitted image features into the normalization layer to obtain a group of optical parameters of the detection object output by the second neural network model.

[0091] The fifth aspect of the embodiment of the present application provides a computer device, comprising a processor, which is used to implement the method for acquiring optical parameters of a detection object provided in the first aspect or the second aspect of the present application when executing a computer program stored on a memory.

[0092] The sixth aspect of the embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is used to implement the method for acquiring optical parameters of a detection object provided in the first aspect or the second aspect of the present application when executing a computer program stored on a memory.

[0093] From the above technical solutions, the embodiments of the present application have the following advantages:

[0094] In the embodiments of the present application, the optical parameters corresponding to the ncs spectral image of the detection object are determined by the neural network model combined with the regression analysis method, so that the automatic output from the ncs spectral image of the detection object to the optical parameters of the detection object is realized. BRIEF DESCRIPTION OF DRAWINGS

[0095] Figure 1 An embodiment of the method for obtaining the optical parameters of the detection object in the embodiments of the present application is shown in the figure;

[0096] Figure 2 An embodiment of the method for obtaining the optical parameters of the detection object in the embodiments of the present application is shown in the figure; Figure 1 The detailed steps of step 103 in the embodiments;

[0097] Figure 3 The structure diagram of the first neural network model in the embodiments of the present application is shown in the figure;

[0098] Figure 4 An embodiment of the method for obtaining the optical parameters of the detection object in the embodiments of the present application is shown in the figure;

[0099] Figure 5 An embodiment of the method for obtaining the optical parameters of the detection object in the embodiments of the present application is shown in the figure;

[0100] Figure 6 The structure diagram of the second neural network model in the embodiments of the present application is shown in the figure; Figure 5 The detailed steps of step 502 in the embodiments;

[0101] Figure 7 The structure diagram of the second neural network model in the embodiments of the present application is shown in the figure;

[0102] Figure 8 An embodiment of the method for obtaining the optical parameters of the detection object in the embodiments of the present application is shown in the figure;

[0103] Figure 9 Another embodiment of the method for obtaining the optical parameters of the detection object in the embodiments of the present application is shown in the figure;

[0104] Figure 10 An embodiment of the device for obtaining the optical parameters of the detection object in the embodiments of the present application is shown in the figure;

[0105] Figure 11 Another embodiment of the device for obtaining the optical parameters of the detection object in the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0106] The embodiment of the present application provides a method for obtaining an optical parameter of a detection object, which can automatically generate the optical parameter of the detection object based on an ncs spectral image of the detection object, thereby improving the convenience from the ncs spectral image to the optical parameter output.

[0107] In order to make the personnel in the technical field better understand the present application scheme, the technical scheme in the embodiment of the present application will be described clearly and completely in the following with reference to the drawings in the embodiment of the present application. Obviously, the described embodiment is only a part of the embodiment of the present application, not all. Based on the embodiment in the present application, all other embodiments obtained by the person skilled in the art without creative labor should belong to the protection scope of the present application.

[0108] The terms "first", "second", "third", "fourth" and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0109] For the convenience of understanding, the method for obtaining the optical parameter of the detection object in the embodiment of the present application is described below, please refer to Figure 1 One embodiment of the method for obtaining the optical parameter of the detection object in the embodiment of the present application comprises:

[0110] 101、obtaining the ncs spectral image of the detection object, wherein the ncs spectral image represents the projection images of the spectral image of the detection object in three mutually orthogonal directions respectively;

[0111] The embodiment of the present application can obtain the ncs spectral image of the detection object through a detection device (such as a photodetector or a photomultiplier tube), wherein the ncs spectral image represents the projection images of the spectral image of the detection object in three mutually orthogonal directions, for example, when the detection object is a wafer, the ncs spectral image of the detection object represents the projection images of the spectral image of the wafer in x, y and z axis directions.

[0112] That is, the ncs spectral image of the detection object includes at least 3 projection images.

[0113] 102、obtaining a plurality of sets of optical parameter values corresponding to the ncs spectral image of the detection object, wherein each set of optical parameters comprises at least two types of optical parameters;

[0114] After obtaining the ncs spectral image of the detection object, a plurality of sets of optical parameter values corresponding to the ncs spectral image of the detection object are obtained, wherein each set of optical parameters comprises at least two types of optical parameters, such as at least two of the thickness d, the bottom width x1, the top width x2, and the azimuth angle θ of the detection object.

[0115] It should be noted here that the plurality of sets of optical parameter values are not actually measured sets of optical parameter values, but sets of optical parameter values generated according to the ncs spectral image of the detection object and the generation function. For example, the first set of optical parameter values includes {d=0.5, x1=1, x1=2, θ=30°}, the second set of optical parameter values includes {d=0.6, x1=2, x1=3, θ=40°}, and the mth set of optical parameter values includes {d=0.9, x1=1, x1=3, θ=60°} and so on.

[0116] 103、inputting the plurality of sets of optical parameter values into a first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model;

[0117] After obtaining the plurality of sets of optical parameter values corresponding to the ncs spectral image of the detection object in step 102, the plurality of sets of optical parameter values are input into a first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model.

[0118] The first neural network model is used to generate an ncs spectral image according to the optical parameter values. The specific model structure of the first neural network model and the process of generating an ncs spectral image by the first neural network model will be described in the embodiments below, and will not be described here.

[0119] 104、determining a target ncs spectral image with the smallest error from the ncs spectral image of the detection object among the plurality of predicted ncs spectral images;

[0120] After obtaining the plurality of predicted ncs spectral images, a target ncs spectral image with the smallest error from the ncs spectral image of the detection object is determined among the plurality of predicted ncs spectral images, that is, a target ncs spectral image most similar to the ncs spectral image of the detection object is determined.

[0121] Specifically, in this embodiment of the application, when determining the target NCS spectral image that is most similar to the NCS spectral image of the detected object from multiple predicted NCS spectral images, the determination can be based on a regression analysis algorithm. The regression analysis algorithm includes linear regression analysis algorithm, ridge regression analysis algorithm, or polynomial regression analysis algorithm, etc. The specific type of regression analysis algorithm is not limited here.

[0122] 105. The target optical parameter group corresponding to the target ncs spectral image is regarded as the optical parameters of the detected object.

[0123] After obtaining the target NCS spectral image, the target optical parameter group corresponding to the target NCS spectral image is regarded as the optical parameters of the object to be detected. For example, assuming that the target optical parameter group is the second group of optical parameters {d = 0.6, x1 = 2, x1 = 3, θ = 40°} in step 102, the optical parameters of this group are regarded as the optical parameters of the object to be detected.

[0124] In this embodiment, the optical parameters corresponding to the NCS spectral image of the detection object are determined by using a neural network model combined with regression analysis, thereby realizing the automated output of the optical parameters of the detection object from the NCS spectral image of the detection object.

[0125] based on Figure 1 The embodiment described above will now be described in detail with reference to step 103. Please refer to [link to relevant documentation]. Figure 2 , Figure 2 Detailed steps for step 103:

[0126] 201. Input the multiple sets of optical parameter values ​​into the convolution module to extract the data features of the multiple sets of optical parameter values;

[0127] The first neural network model in this embodiment includes a convolutional module, a normalization layer, a dropout layer, and a fully connected layer. For ease of understanding, Figure 3 A schematic diagram of the first neural network model is given.

[0128] By inputting multiple sets of optical parameter values ​​into the convolution module, data features such as data distribution and data type can be extracted from these sets of optical parameter values.

[0129] 202. Input the data features into the normalization layer to normalize the data features;

[0130] To prevent excessively large or small data features from being ignored in the model output, a normalization layer is added to the first neural network model in this embodiment to normalize the data features, thereby ensuring that all data features are taken into account in the model output.

[0131] 203. Input the normalized data features into the discard layer to desparse the normalized data features;

[0132] After obtaining the normalized data features, the data features are further input into the discard layer, where the discard layer performs desparsening on the normalized data features.

[0133] Because the normalized data features include multiple 0 values, and 0 values ​​do not play an effective role in the calculation process, the embodiments of this application perform desparsening on the normalized data features. Here, desparsening means deleting 0 values ​​from the normalized data features.

[0134] 204. Input the desparsed data features into the fully connected layer to fit the desparsed data features and obtain the multiple predicted NCS spectral images.

[0135] After obtaining the desparsed data features, the user further inputs the desparsed data features into a fully connected layer. This is because the fully connected layer can fit the desparsed data features, that is, combine the desparsed data features to obtain multiple predicted NCS spectra.

[0136] This application embodiment describes the structure of the first neural network model in detail. The first neural network model includes a normalization layer and a dropout layer. On the one hand, the normalization layer can ensure that the extracted data features are taken into account during the model output process, thereby improving the accuracy of the model output. On the other hand, the dropout layer can delete invalid data in the normalized data features, reduce the computational load of the model, and improve the efficiency of the model output.

[0137] based on Figure 1 The following describes the training process of the first neural network model in the aforementioned embodiment. Please refer to [link / reference]. Figure 4 An embodiment of the training process of the first neural network model in this application includes:

[0138] 401. Obtain a first training sample, the first training sample comprising: multiple NCS spectral images of multiple detection objects and multiple sets of optical parameters corresponding to the multiple NCS spectral images;

[0139] In order to train the first neural network model, it is first necessary to obtain the first training samples, which include multiple NCS spectral images of multiple detection objects and multiple sets of optical parameters corresponding to the multiple NCS spectral images.

[0140] If the plurality of detection objects comprises a plurality of wafers, the plurality of ncs spectral images of the plurality of detection objects and the plurality of sets of optical parameters corresponding to the plurality of ncs spectral images comprise: a plurality of ncs spectral images of the plurality of wafers, and a set of optical parameters corresponding to each wafer ncs spectral image.

[0141] Supposing that the plurality of detection objects comprises m1 wafers, the first training sample data comprises m1 ncs spectral images and m1 sets of optical parameters, and each ncs spectral image corresponds to a set of optical parameters, wherein the optical parameters in the same set at least comprise two types of optical parameters, such as at least two of thickness, bottom width, top width, and azimuth angle.

[0142] 402、inputting the plurality of ncs spectral images of the plurality of detection objects and the plurality of sets of optical parameters corresponding to the plurality of ncs spectral images into the first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model;

[0143] inputting the plurality of ncs spectral images of the plurality of detection objects and the plurality of sets of optical parameters corresponding to the plurality of ncs spectral images into the first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model.

[0144] 403、calculating a loss between the plurality of ncs spectral images of the plurality of detection objects and the plurality of predicted ncs spectral images according to a preset loss function;

[0145] Then, the loss between the plurality of ncs spectral images of the plurality of detection objects and the plurality of predicted ncs spectral images is calculated according to a preset loss function, wherein the preset loss function comprises L1, L2, or cross-entropy loss function, etc., and the type of the loss function is not specifically limited herein.

[0146] 404、training the first neural network model according to the loss and a back propagation algorithm until the loss reaches a target loss.

[0147] After obtaining the loss of the first neural network model, the parameters of the first neural network model are further trained according to the loss and the back propagation algorithm until the loss meets the target loss.

[0148] Specifically, the process of training the first neural network model is a process of continuously updating the model parameters of the first neural network model, and when updating the model parameters of the first neural network model, in order to prevent gradient explosion and improve the convergence speed of the gradient, the learning rate of the parameter update can also be optimized by using an Adamw optimizer when updating the parameters of the first neural network model, and at least has the following advantages:

[0149] 1. Automatically adjust the learning rate of different parameters of the model to better adapt to the changes of the gradient of different parameters;

[0150] 2. The training speed is relatively faster, and it can also better handle sparse gradients;

[0151] 3. The Adamw optimizer performs L2 regularization on larger weight updates, which can improve the training effect of high models.

[0152] The form of the Adamw optimizer itself is consistent with the prior art, and will not be described here.

[0153] In the embodiments of the application, the training process of the first neural network model is described in detail, and when updating the parameters of the first neural network model, the Adamw optimizer is used to optimize the learning rate of parameter update, so that the training speed of the first neural network model is faster and the training effect is better.

[0154] In addition to Figures 1 to 4 In addition to the method for obtaining the optical parameters of the detection object, another method can be used to obtain the optical parameters of the detection object, please refer to Figure 5 An embodiment of obtaining the optical parameters of the detection object in the embodiments of the application includes:

[0155] 501. Obtain the ncs spectral image of the detection object, wherein the ncs spectral image includes the projection images of the spectral image of the detection object in three mutually orthogonal directions respectively;

[0156] The embodiments of the application can obtain the ncs spectral image of the detection object through a detection device (such as a photodetector or a photomultiplier tube), wherein the ncs spectral image represents the projection images of the spectral image of the detection object in three mutually orthogonal directions, for example, when the detection object is a wafer, the ncs spectral image of the detection object represents the projection images of the spectral image of the wafer in the x, y and z axis directions.

[0157] That is, the ncs spectral image of the detection object includes at least three projection images.

[0158] 502. Input the ncs spectral image into the second neural network model to obtain a set of optical parameters of the detection object output by the second neural network model, wherein the second neural network model is used to convert the ncs spectral image of the detection object into a set of optical parameters of the detection object.

[0159] Different from Figures 1 to 4The first neural network model in the embodiment of the present application is used to generate the ncs spectral image of the detection object according to the optical parameter of the detection object, and the second neural network model in the embodiment of the present application can directly generate the optical parameter of the detection object according to the ncs spectral image of the detection object. Therefore, after obtaining the ncs spectral image of the detection object, the embodiment of the present application inputs the spectral image into the second neural network model to obtain a set of optical parameters of the detection object output by the second neural network model.

[0160] The structure of the second neural network model and the process of how the second neural network model outputs the ncs spectral image of the detection object as the corresponding optical parameter will be described in the following embodiments, and will not be described here.

[0161] The embodiment of the present application can convert the ncs spectral image of the detection object into the corresponding optical parameter according to the second neural network model, thereby improving the convenience of obtaining the optical parameter of the detection object.

[0162] Based on Figure 6 The step 502 will be described in detail below with reference to Figure 6 , and Figure 6 The step 502 is a detailed step:

[0163] 601, input the ncs spectral image into the convolution module to extract the image features of the ncs spectral image;

[0164] The second neural network model in the present application includes a convolution module, a dropout layer, a full connection layer and a normalization layer. For the convenience of understanding, Figure 7 The structure diagram of the second neural network model is given.

[0165] When the ncs spectral image is input into the convolution module, the image features of the ncs spectral image can be extracted through the convolution module.

[0166] 602, input the image features into the dropout layer to perform desparsing on the image features;

[0167] Because there may be some 0 values in the image features, and these 0 values are invalid data for the second neural network model, the embodiment of the present application inputs the image features into the dropout layer to perform desparsing on the image features, wherein the operation of desparsing is to delete the 0 values in the image features.

[0168] 603, input the desparsed image features into the full connection layer to fit the desparsed image features;

[0169] After the image features are de-sparse, the de-sparse image features are input into a full connection layer to fit the de-sparse image features, that is, the de-sparse image features are integrated.

[0170] 604、The fitted image features are input into the normalization layer to obtain a set of optical parameters of the detected object output by the second neural network model.

[0171] Finally, the fitted image features are input into the normalization layer to obtain a set of optical parameters of the detected object output by the second neural network model.

[0172] The embodiments of the present application make a detailed description of the structure of the second neural network model, and also make a detailed description of the process of outputting the optical parameters of the detected object by the second neural network model. The dropout layer is arranged in the structure of the second neural network model to delete invalid data in the data features, thereby reducing the operation amount of the model and improving the efficiency and convenience of outputting the optical parameters of the detected object by the second neural network model.

[0173] Based on Figure 1 The embodiments described above, before the ncs spectral images of the detected objects are input into the second neural network model, the second neural network model also needs to be trained. The training process of the second neural network model is described below. Please refer to Figure 8 An embodiment of the training process of the second neural network model in the embodiments of the present application includes:

[0174] 801、Obtain second training sample data, the second training sample data including: an ncs spectral image of each of a plurality of detected objects and a set of optical parameters corresponding to the ncs spectral image;

[0175] Before training the second neural network model, the training sample data needs to be obtained first. Different from Figures 1 to 4 the training sample data of the first neural network model, the training sample data for training the second neural network model is called second training sample data in the embodiments of the present application.

[0176] Specifically, the second training sample data in the embodiments of the present application includes: an ncs spectral image of each of a plurality of detected objects and a set of optical parameters corresponding to the ncs spectral image.

[0177] 802、Input a plurality of ncs spectral images of a plurality of detected objects and a plurality of sets of optical parameters corresponding to the plurality of ncs spectral images into the second neural network model to obtain a plurality of sets of predicted optical parameters output by the second neural network model;

[0178] inputting a plurality of ncs spectral images of a plurality of detection objects and a plurality of sets of optical parameters corresponding to the plurality of ncs spectral images into the second neural network model to obtain a plurality of sets of predicted optical parameters output by the second neural network model, wherein the second neural network model is used to generate corresponding optical parameters from ncs spectral images of detection objects.

[0179] The process of generating a plurality of sets of predicted optical parameters by the second neural network model is similar to Figure 6 Similar to the embodiments described above, details are not repeated here.

[0180] 803、According to the preset loss function, the loss between the plurality of sets of optical parameters of the plurality of detection objects and the plurality of sets of predicted optical parameters is calculated;

[0181] After obtaining the plurality of sets of predicted optical parameters, the loss between the plurality of sets of optical parameters of the plurality of detection objects and the plurality of sets of predicted optical parameters is calculated according to the preset loss function;

[0182] The preset loss function includes L1, L2 or cross-entropy function, etc., which is not limited here.

[0183] 804, according to the loss and back propagation algorithm, the second neural network model is trained until the loss of the second neural network model meets the target loss.

[0184] After obtaining the loss, the second neural network model is further trained according to the loss and back propagation algorithm until the loss of the second neural network model meets the target loss, wherein the process of training the second neural network model is the process of updating the model parameters of the second neural network model.

[0185] Specifically, when updating the parameters of the second neural network model, the learning rate of the parameter update can also be optimized by using Adamw optimizer, and the Adamw optimizer has at least the following advantages:

[0186] 1. Automatically adjust the learning rate of different parameters of the model to better adapt to the change of different parameter gradients;

[0187] 2. The training speed is relatively faster, and it can also better handle sparse gradients;

[0188] 3. The Adamw optimizer performs L2 regularization on the update of larger weights, which can improve the training effect of high models.

[0189] The form of the Adamw optimizer itself is consistent with the prior art, and details are not repeated here.

[0190] In the embodiments of the present application, the training process of the second neural network model is described in detail, and when updating the parameters of the second neural network model, the learning rate of the parameter update is optimized by using the Adamw optimizer, so that the training speed of the second neural network model is faster and the training effect is better.

[0191] Further, based on Figure 8 The embodiments described above, when training the second neural network model, can also use a data parallel method to train the second neural network model, please refer to Figure 9 , Figure 9 Another embodiment of the training process of the second neural network model includes:

[0192] 901, deploy one of the second neural network models on each of the n working nodes, and the initialization parameters of the second neural network models are the same, wherein n is greater than or equal to 2;

[0193] In actual situations, in order to speed up the training of the second neural network model, the second neural network model can be deployed on multiple working nodes (one working node corresponds to one processor or server here), that is, one second neural network model is deployed on each working node, and part of the training samples are used on each working node to train the second neural network model.

[0194] For the convenience of understanding, the following examples are given:

[0195] Assuming that n is 30, 30 second neural network models are deployed on 30 working nodes, that is, one second neural network model is deployed on each working node, and the second neural network models on each working node use the same initialization model parameters.

[0196] 902, divide the training sample data into n groups of sample data;

[0197] Further, the training sample data is divided into n groups of sample data in the embodiments of the present application, assuming that there are 15000 training sample data and n is 30, then the 15000 training sample data are divided into 30 groups, and each group contains 500 sample data.

[0198] 903, update the parameters of the second neural network model on each working node using a group of sample data;

[0199] Further, the parameters of the second neural network model are updated on each working node using a group of sample data.

[0200] The second neural network model deployed on the first worker node is trained by using the first to 500th data, that is, the parameters of the second neural network model are updated.

[0201] 904、after the parameters of the second neural network model are updated m times on each worker node, n updated second neural network models are obtained, 5≤m≤15;

[0202] Specifically, when the second neural network model is trained by using multiple worker nodes, if the parameter average method is used, the parameters of the n updated second neural network models are averaged, but if the model parameters are updated once on each worker node (that is, m=1), and the average of the model parameters in the n worker nodes is calculated, a large network data synchronization overhead and a large calculation overhead are caused, thereby causing a large additional overhead.

[0203] If m is large, the local parameters of each worker node are more diversified, and the model effect after averaging is poor.

[0204] Therefore, m in the embodiment of the present application is greater than or equal to 5 and less than or equal to 15.

[0205] 905、the average of the model parameters of the n updated second neural network models is calculated;

[0206] After the parameters of the second neural network model are updated m times (5≤m≤15) on each worker node, the average of the model parameters of the n updated second neural network models is further calculated, and step 906 is performed according to the average of the model parameters.

[0207] 906、the average of the model parameters is used as the final model parameter of the second neural network model.

[0208] After the average of the model parameters of the second neural network model is obtained, the embodiment of the present application further uses the average of the model parameters as the final model parameter of the second neural network model, thereby improving the training speed of the second neural network model and improving the accuracy of the second neural network model.

[0209] The embodiment of the present application uses a data parallel algorithm to train the same second neural network model on multiple different worker nodes, thereby improving the training speed of the second neural network model while ensuring the accuracy of the second neural network model.

[0210] The method for obtaining the optical parameters of the detection object in the embodiment of the present application is described in detail above, and the device for obtaining the optical parameters of the detection object in the embodiment of the present application is described below. Please refer to Figure 10In an embodiment of the application, the device for obtaining the optical parameters of the detection object comprises:

[0211] The first obtaining unit 1001 is configured to obtain an ncs spectral image of the detection object, wherein the ncs spectral image represents the projection images of the spectral image of the detection object in three mutually orthogonal directions.

[0212] The first obtaining unit 1001 is further configured to obtain a plurality of sets of optical parameter values corresponding to the ncs spectral image of the detection object, wherein each set of optical parameters comprises at least two types of optical parameters.

[0213] The first input / output unit 1002 is configured to input the plurality of sets of optical parameter values into a first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model.

[0214] The first determining unit 1003 is configured to determine, from the plurality of predicted ncs spectral images, a target ncs spectral image having the smallest error with the ncs spectral image of the detection object.

[0215] The first determining unit 1003 is further configured to regard a target set of optical parameters corresponding to the target ncs spectral image as the set of optical parameters of the detection object.

[0216] Preferably, the first determining unit 1003 is specifically configured to:

[0217] perform regression analysis on the plurality of predicted ncs spectral images and the ncs spectral image of the detection object to obtain, from the plurality of predicted ncs spectral images, the target ncs spectral image having the smallest error with the ncs spectral image of the detection object.

[0218] Preferably, the first obtaining unit 1001 is further configured to:

[0219] obtain a first training sample, wherein the first training sample comprises a plurality of ncs spectral images of a plurality of detection objects and a plurality of sets of optical parameters corresponding to the plurality of ncs spectral images.

[0220] The first input / output unit 1001 is further configured to input the plurality of ncs spectral images of the plurality of detection objects and the plurality of sets of optical parameters corresponding to the plurality of ncs spectral images into the first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model.

[0221] Preferably, the device further comprises:

[0222] The computing unit 1004 is configured to calculate a loss between the plurality of ncs spectral images of the plurality of detection objects and the plurality of predicted ncs spectral images according to a preset loss function.

[0223] The first training unit 1005 is configured to train the first neural network model according to the loss and a back propagation algorithm until the loss reaches a target loss.

[0224] Preferably, the first neural network model comprises a convolution module, a normalization layer, a dropout layer and a full connection layer.

[0225] The first input and output unit 1002 is specifically configured to:

[0226] input the plurality of sets of optical parameter values into the convolution module to extract data features of the plurality of sets of optical parameter values;

[0227] input the data features into the normalization layer to normalize the data features;

[0228] input the normalized data features into the dropout layer to de-sparsify the normalized data features;

[0229] input the de-sparsified data features into the full connection layer to fit the de-sparsified data features to obtain the plurality of predicted ncs spectral images.

[0230] In the embodiments of the present application, the first input and output unit 1002 and the first determining unit 1003 combine the neural network model with the regression analysis method to determine the optical parameters corresponding to the ncs spectral images of the detection objects, thereby realizing the automatic output of the optical parameters of the detection objects from the ncs spectral images of the detection objects.

[0231] The embodiments of the present application also provide a device for obtaining optical parameters of detection objects, please refer to Figure 11 Another embodiment of the device for obtaining optical parameters of detection objects in the embodiments of the present application comprises:

[0232] The second obtaining unit 1101 is configured to obtain the ncs spectral images of the detection objects, wherein the ncs spectral images comprise projection images of the spectral images of the detection objects in three mutually orthogonal directions respectively;

[0233] The second input and output unit 1102 is configured to input the ncs spectral images into a second neural network model to obtain a set of optical parameters of the detection objects output by the second neural network model, wherein the second neural network model is configured to convert the ncs spectral images of the detection objects into the set of optical parameters of the detection objects.

[0234] Preferably, the second obtaining unit 1101 is further configured to:

[0235] obtain second training sample data, the second training sample data comprising an ncs spectral image of each detection object in a plurality of detection objects and a group of optical parameters corresponding to the ncs spectral image;

[0236] The apparatus further comprises:

[0237] a second training unit 1103 configured to input the second training sample data into the second neural network model for training until an output of the second neural network model satisfies a target loss.

[0238] Preferably, the second training unit 1103 is specifically configured to:

[0239] input a plurality of ncs spectral images of a plurality of detection objects and a plurality of groups of optical parameters corresponding to the plurality of ncs spectral images into the second neural network model to obtain a plurality of groups of predicted optical parameters output by the second neural network model;

[0240] calculate a loss between the plurality of groups of optical parameters of the plurality of detection objects and the plurality of groups of predicted optical parameters according to a preset loss function;

[0241] train the second neural network model according to the loss and a back propagation algorithm until a loss of the second neural network model satisfies a target loss.

[0242] Preferably, the second training unit 1103 is specifically configured to:

[0243] deploy one of the second neural network models on each of n working nodes, and initialization parameters of the second neural network models are the same, wherein n is greater than or equal to 2;

[0244] divide the training sample data into n groups of sample data evenly;

[0245] update parameters of the second neural network model on each working node by using one group of sample data;

[0246] obtain n updated second neural network models after updating parameters of the second neural network model m times on each working node, 5≤m≤15;

[0247] calculate an average of model parameters of the n updated second neural network models;

[0248] use the average of the model parameters as final model parameters of the second neural network model.

[0249] Preferably, the second neural network model comprises a convolution module, a dropout layer, a full connection layer and a normalization layer;

[0250] The second input and output unit 1102 is specifically configured to:

[0251] input the ncs spectral image into the convolution module to extract image features of the ncs spectral image;

[0252] input the image features into the dropout layer to perform desparsification on the image features;

[0253] input the desparsified image features into the full connection layer to fit the desparsified image features;

[0254] input the fitted image features into the normalization layer to obtain a set of optical parameters of the detection object output by the second neural network model.

[0255] According to the second neural network model, the second input and output unit 1102 converts the ncs spectral image of the detection object into corresponding optical parameters, thereby improving the convenience of obtaining optical parameters of the detection object.

[0256] The above describes the device for obtaining optical parameters of a detection object in the embodiment of the application from the perspective of a modular functional entity. The following describes a computer device in the embodiment of the application from the perspective of hardware processing:

[0257] The computer device is configured to implement the functions of the device for obtaining optical parameters of a detection object. One embodiment of the computer device in the embodiment of the application comprises:

[0258] a processor and a memory;

[0259] The memory is configured to store a computer program, and the processor is configured to execute the computer program stored in the memory, so that the following steps can be implemented:

[0260] obtain an ncs spectral image of the detection object, wherein the ncs spectral image represents projection images of a spectral image of the detection object in three mutually orthogonal directions respectively;

[0261] obtain a plurality of sets of optical parameter values corresponding to the ncs spectral image of the detection object, wherein each set of optical parameters comprises at least two types of optical parameters;

[0262] input the plurality of sets of optical parameter values into a first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model;

[0263] determining a target ncs spectral image with minimum error from the ncs spectral image of the detection object in the plurality of predicted ncs spectral images;

[0264] regarding a target optical parameter group corresponding to the target ncs spectral image as the optical parameter of the detection object.

[0265] In some embodiments of the present application, the processor can further be configured to implement the following steps:

[0266] performing regression analysis on the plurality of predicted ncs spectral images and the ncs spectral image of the detection object to obtain a target ncs spectral image with minimum error from the ncs spectral image of the detection object in the plurality of predicted ncs spectral images.

[0267] In some embodiments of the present application, the processor can further be configured to implement the following steps:

[0268] obtaining a first training sample, the first training sample comprising a plurality of ncs spectral images of a plurality of detection objects and a plurality of optical parameter groups corresponding to the plurality of ncs spectral images;

[0269] inputting the plurality of ncs spectral images of the plurality of detection objects and the plurality of optical parameter groups corresponding to the plurality of ncs spectral images into the first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model;

[0270] calculating a loss between the plurality of ncs spectral images of the plurality of detection objects and the plurality of predicted ncs spectral images according to a preset loss function;

[0271] training the first neural network model according to the loss and a back propagation algorithm until the loss reaches a target loss.

[0272] In some embodiments of the present application, the first neural network model comprises a convolution module, a normalization layer, a dropout layer and a full connection layer, and the processor can further be configured to implement the following steps:

[0273] inputting the plurality of optical parameter values into the convolution module to extract data features of the plurality of optical parameter values;

[0274] inputting the data features into the normalization layer to normalize the data features;

[0275] inputting the normalized data features into the dropout layer to de-sparsify the normalized data features;

[0276] The de-sparse data features are input into the full connection layer to fit the de-sparse data features, so as to obtain the plurality of predicted ncs spectral images.

[0277] The embodiment of the application further provides a computer device, which is also used for realizing the function of the device for acquiring optical parameters of a detection object, and another embodiment of the computer device comprises the following:

[0278] a processor and a memory;

[0279] The memory is used for storing a computer program, and the processor is used for executing the computer program stored in the memory, so that the following steps can be realized:

[0280] The ncs spectral image of the detection object is acquired, and the ncs spectral image comprises projection images of the spectral image of the detection object in three mutually orthogonal directions respectively.

[0281] The ncs spectral image is input into a second neural network model, so as to obtain a group of optical parameters of the detection object output by the second neural network model, wherein the second neural network model is used for converting the ncs spectral image of the detection object into the group of optical parameters of the detection object.

[0282] In some embodiments of the application, before the ncs spectral image is input into the second neural network model, the processor can also be used for realizing the following steps:

[0283] Second training sample data is acquired, and the second training sample data comprises the ncs spectral image of each detection object in a plurality of detection objects and a group of optical parameters corresponding to the ncs spectral image.

[0284] The second training sample data is input into the second neural network model for training until the output of the second neural network model meets a target loss.

[0285] In some embodiments of the application, the processor can also be used for realizing the following steps:

[0286] A plurality of ncs spectral images of a plurality of detection objects and a plurality of groups of optical parameters corresponding to the plurality of ncs spectral images are input into the second neural network model, so as to obtain a plurality of groups of predicted optical parameters output by the second neural network model.

[0287] According to a preset loss function, a loss between the plurality of groups of optical parameters of the plurality of detection objects and the plurality of groups of predicted optical parameters is calculated.

[0288] According to the loss and the back propagation algorithm, the second neural network model is trained until the loss of the second neural network model meets a target loss.

[0289] In some embodiments of the present application, the processor can further be configured to implement the following steps:

[0290] deploying one second neural network model on each of the n working nodes, and the initial parameters of the second neural network models are the same, wherein n is greater than or equal to 2;

[0291] dividing the training sample data into n groups of sample data evenly;

[0292] updating the parameters of the second neural network model on each working node using one group of sample data;

[0293] after updating the parameters of the second neural network model m times on each working node, obtaining n updated second neural network models, 5≤m≤15;

[0294] calculating the average of the model parameters of the n updated second neural network models;

[0295] using the average of the model parameters as the final model parameters of the second neural network model.

[0296] In some embodiments of the present application, the second neural network model comprises a convolution module, a dropout layer, a fully connected layer and a normalization layer; the processor can further be configured to implement the following steps:

[0297] inputting the ncs spectral image into the convolution module to extract image features of the ncs spectral image;

[0298] inputting the image features into the dropout layer to perform de-sparsification on the image features;

[0299] inputting the de-sparsified image features into the fully connected layer to fit the de-sparsified image features;

[0300] inputting the fitted image features into the normalization layer to obtain a group of optical parameters of the detection object output by the second neural network model.

[0301] It can be understood that the processor in the computer device described above can also implement the functions of each unit in the corresponding device embodiments described above when executing the computer program, and details are not repeated here. For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the device for obtaining the optical parameters of the detection object. For example, the computer program can be divided into units in the device for obtaining the optical parameters of the detection object, and each unit can implement the specific functions as described above in the corresponding device for obtaining the optical parameters of the detection object.

[0302] The computer device can be a desktop computer, a notebook computer, a palm computer, a cloud server, and the like. The computer device can include but is not limited to a processor and a memory. Those skilled in the art can understand that the processor and the memory are only examples of the computer device, and do not constitute a limitation on the computer device, and can include more or fewer components, or combine certain components, or different components, for example, the computer device can also include an input / output device, a network access device, a bus, and the like.

[0303] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and the like. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, and the like. The processor is the control center of the computer device, and connects each part of the computer device through various interfaces and lines.

[0304] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the computer device by running or executing the computer program and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required by a function, etc.; and the data storage area can store data created according to the use of the terminal, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0305] The application further provides a computer readable storage medium for realizing the function of the device for acquiring the optical parameters of the detection object, and the computer readable storage medium has a computer program stored thereon, and the computer program is executed by a processor.

[0306] Acquiring an ncs spectral image of the detection object, wherein the ncs spectral image represents the projection images of the spectral image of the detection object in three mutually orthogonal directions respectively;

[0307] Acquiring a plurality of groups of optical parameter values corresponding to the ncs spectral image of the detection object, wherein each group of optical parameters includes at least two types of optical parameters;

[0308] Inputting the plurality of groups of optical parameter values into a first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model;

[0309] Determining a target ncs spectral image with the minimum error from the ncs spectral image of the detection object in the plurality of predicted ncs spectral images;

[0310] Regarding a target group of optical parameters corresponding to the target ncs spectral image as the optical parameters of the detection object.

[0311] In some embodiments of the application, when the computer program is executed by the processor, the processor can further be used to realize the following steps:

[0312] Performing regression analysis on the plurality of predicted ncs spectral images and the ncs spectral image of the detection object to obtain a target ncs spectral image with the minimum error from the ncs spectral image of the detection object in the plurality of predicted ncs spectral images.

[0313] In some embodiments of the present application, when the computer program is executed by the processor, the processor can also be used to implement the following steps:

[0314] The first training sample includes a plurality of ncs spectral images of a plurality of detection objects and a plurality of sets of optical parameters corresponding to the plurality of ncs spectral images;

[0315] The plurality of ncs spectral images of the plurality of detection objects and the plurality of sets of optical parameters corresponding to the plurality of ncs spectral images are input into the first neural network model to obtain a plurality of predicted ncs spectral images output by the first neural network model;

[0316] According to a preset loss function, a loss between the plurality of ncs spectral images of the plurality of detection objects and the plurality of predicted ncs spectral images is calculated;

[0317] According to the loss and a back propagation algorithm, the first neural network model is trained until the loss reaches a target loss.

[0318] In some embodiments of the present application, the first neural network model includes a convolution module, a normalization layer, a dropout layer and a full connection layer, and when the computer program is executed by the processor, the processor can also be used to implement the following steps:

[0319] The plurality of sets of optical parameter values are input into the convolution module to extract data features of the plurality of sets of optical parameter values;

[0320] The data features are input into the normalization layer to normalize the data features;

[0321] The normalized data features are input into the dropout layer to de-sparsify the normalized data features;

[0322] The de-sparsified data features are input into the full connection layer to fit the de-sparsified data features to obtain the plurality of predicted ncs spectral images.

[0323] The present application also provides a computer readable storage medium, which is also used to implement the function of the device for obtaining optical parameters of a detection object, and has a computer program stored thereon. When the computer program is executed by the processor, the processor can be used to execute the following steps:

[0324] The ncs spectral image of the detection object includes projection images of the spectral image of the detection object in three mutually orthogonal directions respectively;

[0325] inputting the ncs spectral image into a second neural network model to obtain a set of optical parameters of the detection object output by the second neural network model, wherein the second neural network model is configured to convert the ncs spectral image of the detection object into a set of optical parameters of the detection object.

[0326] In some embodiments of the present application, before the ncs spectral image is input into the second neural network model, the processor, when executing the computer program, can further be configured to implement the following steps:

[0327] obtaining second training sample data, wherein the second training sample data comprises an ncs spectral image of each of a plurality of detection objects and a set of optical parameters corresponding to the ncs spectral image;

[0328] inputting the second training sample data into the second neural network model for training until the output of the second neural network model meets a target loss.

[0329] In some embodiments of the present application, the processor, when executing the computer program, can further be configured to implement the following steps:

[0330] inputting a plurality of ncs spectral images of a plurality of detection objects and a plurality of sets of optical parameters corresponding to the plurality of ncs spectral images into the second neural network model to obtain a plurality of sets of predicted optical parameters output by the second neural network model;

[0331] calculating a loss between the plurality of sets of optical parameters of the plurality of detection objects and the plurality of sets of predicted optical parameters according to a preset loss function;

[0332] training the second neural network model according to the loss and a back propagation algorithm until the loss of the second neural network model meets a target loss.

[0333] In some embodiments of the present application, the processor, when executing the computer program, can further be configured to implement the following steps:

[0334] deploying one second neural network model on each of n working nodes, and the initialization parameters of the second neural network models are the same, wherein n is greater than or equal to 2;

[0335] dividing the training sample data into n groups of sample data evenly;

[0336] updating the parameters of the second neural network model on each working node by using a group of sample data;

[0337] updating parameters of the second neural network model m times on each worker node, wherein m is an integer between 5 and 15, and obtaining n updated second neural network models;

[0338] calculating an average of model parameters of the n updated second neural network models;

[0339] using the average of the model parameters as final model parameters of the second neural network model.

[0340] In some embodiments of the present application, the second neural network model comprises a convolution module, a dropout layer, a fully connected layer and a normalization layer; when the computer program is executed by the processor, the processor can also be used to implement the following steps:

[0341] inputting the ncs spectral image into the convolution module to extract image features of the ncs spectral image;

[0342] inputting the image features into the dropout layer to perform desparsing on the image features;

[0343] inputting the desparsed image features into the fully connected layer to fit the desparsed image features;

[0344] inputting the fitted image features into the normalization layer to obtain a set of optical parameters of the detection object output by the second neural network model.

[0345] It can be understood that the integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a corresponding computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned corresponding embodiment methods of the present application can also be completed by instructing related hardware through a computer program, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the contents included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0346] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. In actual implementation, another division mode can be adopted. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0347] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0348] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software function unit.

[0349] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not to limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for acquiring optical parameters of a detection object, characterized in that, include: Obtain the NCS spectral image of the object to be detected, wherein the NCS spectral image represents the projection image of the spectral image of the object to be detected in three mutually orthogonal directions; Obtain multiple sets of optical parameter values ​​corresponding to the NCS spectral image of the detected object, wherein each set of optical parameters includes at least two types of optical parameters; The multiple sets of optical parameter values ​​are input into the first neural network model to obtain multiple predicted ncs spectral images output by the first neural network model. Among the multiple predicted NCS spectral images, the target NCS spectral image with the smallest error compared to the NCS spectral image of the detected object is determined. The target optical parameter set corresponding to the target NCS spectral image is regarded as the optical parameter set of the detected object; Before inputting the multiple sets of optical parameter values ​​into the first neural network model, the method further includes: A first training sample is obtained, which includes: multiple NCS spectral images of multiple detection objects and multiple sets of optical parameters corresponding to the multiple NCS spectral images; Multiple NCS spectral images of the multiple detection objects and multiple sets of optical parameters corresponding to the multiple NCS spectral images are input into the first neural network model to obtain multiple predicted NCS spectral images output by the first neural network model; Based on a preset loss function, calculate the loss between multiple NCS spectral images of multiple detected objects and the multiple predicted NCS spectral images; The parameters of the first neural network model are trained according to the loss and backpropagation algorithm until the loss reaches the target loss.

2. The method according to claim 1, characterized in that, Among the multiple predicted NCS spectral images, determining the target NCS spectral image with the smallest error compared to the NCS spectral image of the detected object includes: Regression analysis is performed on the multiple predicted NCS spectral images and the NCS spectral image of the detected object to obtain the target NCS spectral image with the smallest error compared to the NCS spectral image of the detected object among the multiple predicted NCS spectral images.

3. The method according to claim 1, characterized in that, The first neural network model includes: a convolutional module, a normalization layer, a dropout layer, and a fully connected layer; The multiple sets of optical parameter values ​​are input into a first neural network model to obtain multiple predicted NCS spectral images output by the first neural network model, including: The multiple sets of optical parameter values ​​are input into the convolution module to extract the data features of the multiple sets of optical parameter values; The data features are input into the normalization layer to normalize the data features; The normalized data features are input into the discard layer to desparse the normalized data features; The desparsed data features are input into the fully connected layer to fit the desparsed data features and obtain the multiple predicted NCS spectral images.

4. A method for acquiring optical parameters of a detection object, characterized in that, include: Acquire the NCS spectral image of the object to be detected, wherein the NCS spectral image includes the projection images of the spectral image of the object to be detected in three mutually orthogonal directions; The NCS spectral image is input into a second neural network model to obtain a set of optical parameters of the detected object output by the second neural network model, wherein the second neural network model is used to convert the NCS spectral image of the detected object into a set of optical parameters of the detected object; Before inputting the ncs spectral image into the second neural network model, the method further includes: Acquire second training sample data, which includes: an NCS spectral image of each of the multiple detection objects and a set of optical parameters corresponding to the NCS spectral image; The second training sample data is input into the second neural network model for training until the output of the second neural network model meets the target loss.

5. The method according to claim 4, characterized in that, The step of inputting the second training sample data into the second neural network model for training includes: Multiple NCS spectral images of multiple detection objects and multiple sets of optical parameters corresponding to the multiple NCS spectral images are input into the second neural network model to obtain multiple sets of predicted optical parameters output by the second neural network model; Based on a preset loss function, calculate the loss between multiple sets of optical parameters of the multiple detection objects and the multiple sets of predicted optical parameters; The second neural network model is trained according to the loss and backpropagation algorithm until the loss of the second neural network model meets the target loss.

6. The method according to claim 4, characterized in that, The second training sample data is input into the second neural network model for training, including: Deploy one of the second neural network models on each of the n working nodes, and the initialization parameters of the second neural network models are the same, where n is greater than or equal to 2; The second training sample data is divided into n groups of sample data on an average basis; At each working node, the parameters of the second neural network model are updated using a set of sample data; After updating the parameters of the second neural network model m times at each working node, n updated second neural network models are obtained, where 5≤m≤15; Calculate the average of the model parameters of the n updated second neural network models; The average value of the model parameters is used as the final model parameters of the second neural network model.

7. The method according to claim 4, characterized in that, The second neural network model includes convolutional modules, dropout layers, fully connected layers, and normalization layers; The NCS spectral image is input into a second neural network model to obtain a set of optical parameters of the detected object output by the second neural network model, including: The NCS spectral image is input into the convolution module to extract the image features of the NCS spectral image; The image features are input into the discard layer to perform desparsing on the image features; The desparsed image features are input into the fully connected layer to fit the desparsed image features; The fitted image features are input into the normalization layer to obtain a set of optical parameters of the detected object output by the second neural network model.

8. A device for acquiring optical parameters of a detection object, characterized in that, include: The first acquisition unit is used to acquire the NCS spectral image of the object to be detected, wherein the NCS spectral image represents the projection images of the spectral image of the object to be detected in three mutually orthogonal directions; The first acquisition unit is further configured to acquire multiple sets of optical parameter values ​​corresponding to the NCS spectral image of the detection object, wherein each set of optical parameters includes at least two types of optical parameters; The first input-output unit is used to input the multiple sets of optical parameter values ​​into the first neural network model to obtain multiple predicted NCS spectral images output by the first neural network model. The first determining unit is configured to determine, among the plurality of predicted NCS spectral images, the target NCS spectral image with the smallest error to the NCS spectral image of the detected object; The first determining unit is further configured to regard the target optical parameter group corresponding to the target ncs spectral image as the optical parameter group of the detection object; The first acquisition unit is further configured to: A first training sample is obtained, which includes: multiple NCS spectral images of multiple detection objects and multiple sets of optical parameters corresponding to the multiple NCS spectral images; The first input / output unit is further configured to input multiple NCS spectral images of the multiple detection objects and multiple sets of optical parameters corresponding to the multiple NCS spectral images into the first neural network model to obtain multiple predicted NCS spectral images output by the first neural network model; The device further includes: The calculation unit is used to calculate the loss between multiple NCS spectral images of multiple detected objects and the multiple predicted NCS spectral images according to a preset loss function; The training unit is used to train the first neural network model according to the loss and the backpropagation algorithm until the loss reaches the target loss.

9. A device for acquiring optical parameters of a detection object, characterized in that, include: The second acquisition unit is used to acquire the NCS spectral image of the object being detected, wherein the NCS spectral image includes projection images of the spectral image of the object being detected in three mutually orthogonal directions. The second input / output unit is used to input the NCS spectral image into the second neural network model to obtain a set of optical parameters of the detected object output by the second neural network model, wherein the second neural network model is used to convert the NCS spectral image of the detected object into a set of optical parameters of the detected object; The second acquisition unit is also used for: Acquire second training sample data, which includes: an NCS spectral image of each of the multiple detection objects and a set of optical parameters corresponding to the NCS spectral image; The device further includes: The training unit is used to input the second training sample data into the second neural network model for training until the output of the second neural network model meets the target loss.

10. A computer device, comprising a processor, characterized in that, When the processor executes a computer program stored in the memory, it is used to implement the method for obtaining optical parameters of the object to be detected as described in any one of claims 1 to 3, or the method for obtaining optical parameters of the object to be detected as described in any one of claims 4 to 7.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it is used to implement the method for obtaining optical parameters of the object to be detected as described in any one of claims 1 to 3, or the method for obtaining optical parameters of the object to be detected as described in any one of claims 4 to 7.

Images