A method, device, electronic device and storage medium for zircon age prediction

Abstract

The present invention provides a method, an apparatus, an electronic device and a storage medium for predicting the age of zircon. The method includes: collecting and organizing the cathodoluminescence (CL) images of zircon, where the zircon characteristic information in the CL images at least includes the growth zoning characteristics of zircon; performing prediction processing on the zircon CL images by using a preset target zircon age prediction model, and obtaining an age prediction result of zircon; the target zircon age prediction model includes a zoning extraction module, a multi-scale feature fusion module and an age prediction module, the zoning extraction module is used for extracting the growth zoning characteristics in the CL image information, the multi-scale feature fusion module is used for performing multi-scale feature extraction and fusion on the growth zoning characteristics to obtain multi-scale spatial information, and the age prediction module is used for obtaining a final zircon age prediction result based on the multi-scale spatial information.

Description

Technical Field

[0001] The present invention belongs to the technical fields of artificial intelligence and geoscience big data, and particularly relates to a method, device, electronic device and storage medium for predicting zircon age. Background Art

[0002] Image age prediction is an important task in the field of computer vision. In the current era of continuous progress in sedimentary geology, more and more computer vision methods are being introduced into the field of geology to solve the problems of efficiency and accuracy of traditional research methods. In geological research, a large amount of data information, especially microscopic image data of minerals, provides rich information for scientists to study the generation, evolution and environmental conditions of minerals. In zircon research, the age, generation environment and evolution characteristics of zircon are important clues for geologists to study sedimentary history and tectonic movements. Rapid and accurate prediction of zircon age is the core research direction that geologists have long been concerned about.

[0003] Among them, traditional zircon age analysis methods rely on microscope observation and element ratio analysis, such as U-Pb determination, etc. These methods are not only time-consuming and laborious, but also require complex experimental equipment and professional background support. At the same time, they are easily affected by factors such as sample preservation status and analysis conditions, resulting in inaccurate age analysis. And there is a correlation between zircon CL (cathodoluminescence) images and age prediction. The growth zone width, quantity and texture characteristics formed by zircon in different geological history stages are closely related to the age of zircon; however, there is a lack of a method in the existing related technologies that can accurately predict the zircon age through zircon CL images.

[0004] Therefore, the present application anticipates a method for predicting zircon age based on annulus extraction technology. Summary of the Invention

[0005] Aiming at the problems existing in the prior art, the present invention provides a method, device, electronic device and storage medium for predicting zircon age, which can efficiently and accurately predict the zircon age.

[0006] In a first aspect, an embodiment of the present invention provides a method for predicting zircon age, including:

[0007] Collect and sort out the cathodoluminescence CL images of zircon, and the zircon feature information in the CL images at least includes the growth zone characteristics of zircon;

[0008] Use a preset target zircon age prediction model to perform prediction processing on the zircon CL image, and obtain the age prediction result of the zircon;

[0009] The target zircon age prediction model includes a zoning extraction module, a multi-scale feature fusion module, and an age prediction module. The zoning extraction module is used to extract the growth zoning features in the CL image information. The multi-scale feature fusion module is used to perform multi-scale feature extraction and fusion on the growth zoning features to obtain multi-scale spatial information. The age prediction module is used to obtain the final zircon age prediction result based on the multi-scale spatial information.

[0010] Further, before the target zircon age prediction model predicts the age of zircon, it also includes training and obtaining the target zircon age model. The training and obtaining of the target zircon age model includes:

[0011] Collect and organize the CL images of the zircon to be tested and the age labels of the zircon to be tested. The zircon feature information in the CL image includes at least the growth zoning features of the zircon.

[0012] Make training datasets, validation datasets, and test datasets from the CL images of the zircon to be tested and the age labels.

[0013] Construct an initial zircon age prediction model.

[0014] Based on the training dataset and the validation dataset, use the stochastic gradient descent method to optimize the initial zircon age prediction model to obtain an optimized zircon age prediction model.

[0015] Use the optimized zircon age prediction model to predict the CL images of zircon in the test dataset to obtain the predicted age results.

[0016] Based on the comparison and analysis of the predicted age results and the corresponding true age labels in the test dataset, obtain the target zircon age prediction model.

[0017] Further, obtaining the optimized zircon age prediction model includes:

[0018] Use the stochastic gradient descent optimizer to train the initial zircon age prediction model; and

[0019] Use the data augmentation strategy to optimize the initial zircon age prediction model.

[0020] Further, input the CL image into the target zircon age prediction model according to a preset size.

[0021] In the zircon zoning extraction module, the CL image of the zircon is input, the image is cropped and normalized according to a preset size to obtain segmented image patches, the image patches are input into the structure of the ViT model, processed layer by layer, the feature relationship is calculated through the self-attention mechanism, and the output features of the ViT model are mixed and calculated through a convolution and a multi-layer perceptron module to obtain growth zoning features;

[0022] In the multi-scale feature fusion module, the growth zoning features are extracted layer by layer through a transformer block, the correlation between local and global features is calculated through the self-attention mechanism, the feature map is non-linearly transformed through a feed-forward neural network, different-scale image patches are integrated through patch merging, and the output features of different layers are spliced and fused to integrate multi-scale spatial information;

[0023] In the age prediction module, the time series information of the multi-scale spatial information is captured through a long short-term memory network, a multi-layer perceptron is used for dimensionality reduction and non-linear mapping to generate an initial age prediction result, the dual sorting loss is used to optimize the predicted value, the relative time order between samples is sorted, and the final zircon age prediction result is output.

[0024] Further, obtaining the growth zoning features includes:

[0025]

[0026] F out = MLP(Conv(F Attention , W conv ));

[0027] In the formula, where Q, K, and V represent the query vector, key vector, and value vector respectively, d k is the dimension of the key, F Attention is the output feature of the self-attention mechanism, and W conv is the convolution weight.

[0028] Further, obtaining the multi-scale spatial information includes:

[0029] F concat = [F trans1 , F trans2 , F trans3 , F trans4 ;

[0030] F unified = MLP(F concat );

[0031] F up = Upsample(F unified );

[0032] In the formula, F concat is the feature map obtained by splicing features through F trans1 , F trans2 , F trans3 , F trans4 The feature map obtained by splicing features, F unified is the multi-scale feature fusion feature, F up is the high-resolution feature map after upsampling.

[0033] Furthermore, the final zircon age prediction result is obtained, including:

[0034] F time = LSTM(F up );

[0035] Age pred = MLP(F time );

[0036]

[0037] In the formula, F up is the feature map after multi-scale feature fusion, F time is the time series feature generated by LSTM, Age pred is the age prediction value, and the prediction value is optimized using Pairwise RankingLoss.

[0038] In a second aspect, an embodiment of the present disclosure provides a device for zircon age prediction, and the device includes:

[0039] An acquisition unit, configured to collect and organize the cathodoluminescence (CL) image of zircon, and the zircon feature information in the CL image at least includes the growth zoning feature of zircon;

[0040] A zircon age prediction unit, configured to perform prediction processing on the zircon CL image by using a preset target zircon age prediction model, and obtain the age prediction result of zircon; wherein, the target zircon age prediction model includes an annulus extraction module, a multi-scale feature fusion module, and an age prediction module, the annulus extraction module is used to extract the growth annulus feature in the CL image information, the multi-scale feature fusion module is used to perform multi-scale feature extraction and fusion on the growth annulus feature to obtain multi-scale spatial information, and the age prediction module is used to obtain the final zircon age prediction result based on the multi-scale spatial information.

[0041] In a third aspect, an embodiment of the present disclosure provides an electronic device, and the electronic device includes:

[0042] At least one processor; and,

[0043] A memory communicatively connected to the at least one processor; wherein,

[0044] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method for zircon age prediction described above.

[0045] In a fourth aspect, an embodiment of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the method for zircon age prediction described above.

[0046] Some of the other optional features and technical effects of the embodiments of the present invention are described below, and some can be understood by reading this article.

[0047] Compared with the prior art, the present invention has the following beneficial technical effects:

[0048] The present invention provides a method for zircon age prediction, including: collecting and sorting out the cathodoluminescence (CL) images of zircons, where the zircon feature information in the CL images at least includes the growth zoning features of zircons; performing prediction processing on the zircon CL images using a preset target zircon age prediction model to obtain an age prediction result of the zircons; the target zircon age prediction model includes a zoning extraction module, a multi-scale feature fusion module, and an age prediction module, the zoning extraction module is used to extract the growth zoning features in the CL image information, the multi-scale feature fusion module is used to perform multi-scale feature extraction and fusion on the growth zoning features to obtain multi-scale spatial information, and the age prediction module is used to obtain the final zircon age prediction result based on the multi-scale spatial information. By feeding the zircon CL images into the target zircon age prediction model, this method can automatically identify the zircon age, which has efficient and accurate prediction capabilities. Compared with the prediction results of other prediction models, this application only needs to perform simple resolution adaptation to predict the zircon age. Therefore, this application has wide applications in geological image age prediction. Description of the Drawings

[0049] Figure 1 Shows a schematic flowchart of a method for zircon age prediction according to an embodiment of the present disclosure;

[0050] Figure 2 Shows a schematic diagram of a zircon age prediction model according to an embodiment of the present disclosure;

[0051] Figure 3 Shows some results of zircon age prediction according to an embodiment of the present disclosure;

[0052] Figure 4Shows a device diagram for zircon age prediction according to an embodiment of the present disclosure. Detailed implementation manners

[0053] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0054] The following uses specific specific examples to illustrate the implementation manners of the present disclosure. Those skilled in the art can easily understand other advantages and effects of the present disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all the embodiments. The present disclosure can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present disclosure without creative efforts belong to the scope of protection of the present disclosure.

[0055] It should be noted that the following describes various aspects of the embodiments within the scope of the appended claims. It should be obvious that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is illustrative only. Based on the present disclosure, those skilled in the art should understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, this device and / or this method can be implemented using other structures and / or functions in addition to one or more of the aspects described herein.

[0056] It should also be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present disclosure in a schematic manner. The diagrams only show the components related to the present disclosure and are not drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in its actual implementation can be an arbitrary change, and the component layout type may also be more complex.

[0057] In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0058] In the first aspect, Figure 1 Shows a method 100 for zircon age prediction provided by an embodiment of the present invention, including the following steps:

[0059] At step S101, collect and organize the cathodoluminescence (CL) images of zircons. The zircon characteristic information in the CL images includes at least the growth zoning characteristics of zircons.

[0060] Specifically, the growth zoning of zircons is formed when they crystallize in magma, usually manifested as a series of alternating bright and dark bands. The presence and characteristics of these zones can reflect important information such as changes in magma composition, crystallization processes, and the multi - periodicity of magmatic activities.

[0061] Next, proceed to step S102.

[0062] At step S102, use a preset target zircon age prediction model to perform prediction processing on the zircon CL images and obtain the zircon age prediction result. Among them, the target zircon age prediction model includes a zoning extraction module, a multi - scale feature fusion module, and an age prediction module. The zoning extraction module is used to extract the growth zoning characteristics from the CL image information. The multi - scale feature fusion module is used to perform multi - scale feature extraction and fusion on the growth zoning characteristics to obtain multi - scale spatial information. The age prediction module is used to obtain the final zircon age prediction result based on the multi - scale spatial information.

[0063] In some embodiments, before the target zircon age prediction model predicts the age of zircons, it further includes training and obtaining the target zircon age model. The training and obtaining of the target zircon age model includes:

[0064] Collect and organize the CL images of zircons to be tested and the age labels of zircons to be tested. The zircon characteristic information in the CL images includes at least the growth zoning characteristics of zircons.

[0065] Make training datasets, validation datasets, and test datasets from the CL images of zircons to be tested and the age labels.

[0066] Construct an initial zircon age prediction model.

[0067] Based on the training dataset and the validation dataset, use the stochastic gradient descent method to optimize the initial zircon age prediction model to obtain an optimized zircon age prediction model.

[0068] Use the optimized zircon age prediction model to predict the zircon CL images in the test dataset to obtain the predicted age results.

[0069] Based on the comparison and analysis of the predicted age results and the corresponding true age labels in the test dataset, obtain the target zircon age prediction model.

[0070] It should be noted that in the embodiments of the present disclosure, the age label and the annual zero error are used in the training optimization and evaluation index stages of the model. The age label is the true zircon age and is used as the target value in supervised learning. When using the Pairwise Ranking Loss to optimize the predicted value of the model, the sorting relationship of the model is evaluated according to the age label. The age error is used to evaluate the accuracy of the model.

[0071] Specifically, Table 1 shows the zircon CL image dataset of the embodiments of the present disclosure, including the total number of training datasets (TrainDataset), validation datasets (Valid Dataset), test datasets (Test Dataset), and images (Image). Among them, the training dataset is 1324 groups, the validation dataset is 166 groups, the test dataset is 166 groups, and the total number of images is 1656 groups.

[0072] Table 1. Zircon CL Image Dataset

[0073]

[0074] Further, in some embodiments, obtaining the optimized zircon age prediction model includes:

[0075] Training the initial zircon age prediction model using a stochastic gradient descent optimizer; and

[0076] Optimizing the initial zircon age prediction model using a data augmentation strategy.

[0077] Specifically, data augmentation (Augment) is performed after the input image (Image). In some embodiments, the data augmentation strategy includes random cropping, resizing, horizontal flipping, random rotation, color adjustment, and an auto-augmentation strategy (AutoAugment).

[0078] The stochastic gradient descent optimizer (SGD) is an iterative method for optimizing model parameters. By updating the parameters using only a small portion of the training samples in each iteration, the training speed is accelerated and it helps to jump out of local optima. During training, the data augmentation strategy is applied in real time to generate augmented training samples, and these samples are fed to the SGD optimizer for parameter update. This helps the model learn more robust feature representations over a wider data distribution.

[0079] In some embodiments, the CL image is input into the target zircon age prediction model according to a preset size.

[0080] In the zircon zoning extraction module, the CL image of the zircon is input, the image is cropped and normalized according to a preset size to obtain segmented image patches, and the image patches are input into the structure of the ViT model, processed layer by layer, the feature relationship is calculated through the self-attention mechanism, and the output features of the ViT model are mixed and calculated through the convolution and multi-layer perceptron modules to obtain the growth zoning features.

[0081] In the multi-scale feature fusion module, the growth zoning features are extracted layer by layer through the transformer block, the correlation between local and global features is calculated through the self-attention mechanism, the feature map is non-linearly transformed through the feed-forward neural network, the image patches of different scales are integrated through patch merging, and the output features of different layers are spliced and fused to integrate multi-scale spatial information.

[0082] In the age prediction module, the time series information of the multi-scale spatial information is captured through the long short-term memory network, the multi-layer perceptron is used for dimensionality reduction and non-linear mapping to generate the initial age prediction result, and the dual ranking loss is used to optimize the predicted value to rank the relative time order between samples, and the final zircon age prediction result is output.

[0083] Specifically, obtaining the growth zoning features includes:

[0084]

[0085] F out = MLP(Conv(F Attention , W conv ));

[0086] In the formula, Q, K, and V respectively represent the query vector, key vector, and value vector, d k is the dimension of the key, F Attention is the output feature of the self-attention mechanism, and W conv is the convolution weight.

[0087] Specifically, obtaining the multi-scale spatial information includes:

[0088] F concat = [F trans1 , F trans2 , F trans3 , F trans4 ;

[0089] F unified = MLP(F concat );

[0090] F up = Upsample(F unified );

[0091] In the formula, F concat is the feature map obtained by feature splicing through F trans1 , F trans2 , F trans3 , F trans4 , and F unified is the multi-scale feature fusion feature, and F up is the high-resolution feature map after upsampling.

[0092] Specifically, obtaining the final zircon age prediction result includes:

[0093] F time = LSTM(F up );

[0094] Age pred = MLP(F time );

[0095]

[0096] In the formula, F up is the feature map after multi-scale feature fusion, F time is the time series feature generated by LSTM, and Age pred is the age prediction value, and the prediction value is optimized using Pairwise RankingLoss.

[0097] Furthermore, Figure 2 shows a schematic diagram of a target zircon age prediction model in an embodiment of the present disclosure. As Figure 2 shown, in the embodiment of the present disclosure, the zircon CL image (Image) is input into the target zircon age prediction model according to a preset size, that is, according to the input image of a given size of H×W.

[0098] In the zircon annulus extraction module (Zones Extract Model), the CL image of the zircon is input, the image is cropped and normalized according to the given size of 224×224, and the segmented patches are input into the structure of VisionTransformer (ViT). The Vision Transformer (ViT) structure is a deep learning model that applies the Transformer architecture to visual recognition tasks; the feature relationship is calculated through the self-attention mechanism (Self-attention), and the output features are mixed and calculated through a 3×3 convolution and an MLP multi-layer perceptron module (MLP, Multi-LayerPerceptron) to extract the growth annulus features.

[0099] In the multi-scale feature fusion module (Multi-scale Feature Fusion Module), first, the segmented patches are converted into vector form through overlap patch embeddings, and then features are extracted layer by layer through the TransformerBlock transformer block. Specifically, there are multiple Transformer Block transformer blocks, including Transformer Block1, TransformerBlock2, Transformer Block3, and Transformer Block4. Each Transformer Block transformer block includes an efficient self-attention mechanism ESA (Efficient Self-Attention), a feed-forward neural network FFN (Feed-Forward Network), and patch merging. The efficient self-attention mechanism ESA (Efficient Self-Attention) calculates the correlation between local and global features. The feed-forward neural network FFN (Feed-Forward Network) performs non-linear transformation on the feature map. The patch merging integrates patches of different scales. Finally, the output features of different layers are concatenated and fused to integrate multi-scale spatial information.

[0100] In the age prediction module (Age PredictionModule), the long short-term memory network LSTM (Long Short-Term Memory) is used to capture the time series information of the multi-scale features of zircon zoning. The multi-layer perceptron MLP (Multi-LayerPerceptron) is used for dimensionality reduction and non-linear mapping to generate the age prediction result. The pairwise ranking loss is used to optimize the predicted value to meet the relative time order between samples, and the final age prediction value of zircon is output.

[0101] Exemplarily, Table 2 shows the results of predicting the zircon age based on the zircon CL image by the target zircon age prediction model provided in this embodiment, and compares it with other models to analyze the improvement of the accuracy ACC@0, ACC@1, ACC@5, and ACC@10. Among them, other models include the model Resnet50, coral, and the model MiVOLO.

[0102] Table 2. Comparison of zircon age prediction results

[0103]

[0104] It should be noted that Predict age(%) is the age error, which is an indicator of the accuracy of age prediction and is usually used to evaluate the performance of an age prediction model. It represents the percentage of the predicted age by the model that matches the actual age. ResNet50 is a deep convolutional neural network model. MiVOLO is a multi-input transformer for age and gender estimation. The Coral framework (CorrelationAlignment) is an unsupervised transfer learning algorithm. The ACC@0 is the proportion of completely correct predictions by the model, and ACC@1 represents the proportion of the category with the highest predicted probability by the model that matches the actual category. ACC@5 represents the proportion that the true label is within the top 5 categories with the highest predicted probabilities by the model. ACC@10 represents the proportion that the true label is within the top 10 categories with the highest predicted probabilities by the model.

[0105] Exemplarily, Figure 3 The zircon age prediction result diagram of the embodiment of the present disclosure is shown. The diagram shows a zircon image, a zircon CL image, an age label, an age error (sigma), and an age prediction result (predict age). It can be seen from the diagram that by using the target zircon age prediction model provided in this embodiment, the age of zircon can be predicted more accurately.

[0106] The second embodiment of the present invention also provides a device for zircon age prediction. The device includes:

[0107] An acquisition unit configured to collect and organize the cathodoluminescence (CL) image of zircon, and the zircon feature information in the CL image at least includes the growth zoning feature of zircon;

[0108] A zircon age prediction unit configured to perform prediction processing on the zircon CL image by using a preset target zircon age prediction model and obtain an age prediction result of zircon; wherein, the target zircon age prediction model includes an annulus extraction module, a multi-scale feature fusion module, and an age prediction module. The annulus extraction module is used to extract the growth zoning feature in the CL image information, the multi-scale feature fusion module is used to perform multi-scale feature extraction and fusion on the growth zoning feature to obtain multi-scale spatial information, and the age prediction module is used to obtain the final zircon age prediction result based on the multi-scale spatial information.

[0109] The third embodiment of the present invention also provides an electronic device, which includes:

[0110] At least one processor; and,

[0111] A memory communicatively connected to the at least one processor; wherein,

[0112] The memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the method for zircon age prediction according to any one of the foregoing embodiments.

[0113] The fourth embodiment of the present invention further provides a non-transitory computer-readable storage medium, and the non-transitory computer-readable storage medium stores computer instructions for causing the computer to execute the method for zircon age prediction according to any one of the foregoing embodiments.

[0114] The fifth embodiment of the present invention further provides a computer program product, and the computer program product includes a computing program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer is caused to execute the method for zircon age prediction according to any one of the foregoing embodiments.

[0115] Figure 4 A schematic diagram showing a method that can implement the embodiments of the present invention or a device 1000 that implements the embodiments of the present invention. In some embodiments, it may include more or fewer devices than shown. In some embodiments, it can be implemented using a single or multiple devices. In some embodiments, it can be implemented using cloud or distributed devices.

[0116] As Figure 4 shown, the device 1000 includes a processor 1001, which can perform various appropriate operations and processes according to the programs and / or data stored in the read-only memory (ROM) 1002 or the programs and / or data loaded from the storage section 1008 into the random access memory (RAM) 1003. The processor 1001 can be a multi-core processor or can include multiple processors. In some embodiments, the processor 1001 can include a general main processor and one or more special coprocessors, for example, a central processing unit (CPU), a graphics processing unit (GPU), a neural network processing unit (NPU), a digital signal processor (DSP), and so on. In the RAM 1003, various programs and data required for the operation of the device 1000 are also stored. The processor 1001, the ROM 1002, and the RAM 1003 are connected to each other through a bus 1004. The input / output (I / O) interface 1005 is also connected to the bus 1004.

[0117] The above-mentioned processor and the memory are jointly used to execute a program stored in the memory. When the program is executed by a computer, it can implement the methods, steps or functions described in the above embodiments.

[0118] The following components are connected to the I / O interface 1005: an input section 1006 including a keyboard, a mouse, a touch screen, etc.; an output section 1007 including such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN card, a modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as required. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1010 as required, so that a computer program read from it can be installed into the storage section 1008 as required. Figure 4 Only some components are schematically shown, and it does not mean that the device 1000 only includes Figure 4 the components shown.

[0119] The systems, devices, modules or units illustrated in the above embodiments can be implemented by a computer or its associated components. The computer can be, for example, a mobile terminal, a smart phone, a personal computer, a laptop computer, an in-vehicle human-machine interaction device, a personal digital assistant, a media player, a navigation device, a game console, a tablet computer, a wearable device, a smart TV, an Internet of Things system, a smart home, an industrial computer, a server or a combination thereof.

[0120] Although not shown, in an embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program / instructions are stored. When the computer program / instructions are executed by a processor, the method for zircon age prediction described in the embodiment is implemented.

[0121] The storage medium in the embodiment of the present invention includes permanent and non-permanent, removable and non-removable articles that can implement information storage by any method or technology. Examples of storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information accessible by a computing device.

[0122] Although not shown, an embodiment of the present invention also provides a computer program product, including: a computer program / instructions, which when executed by a processor, implement the method for zircon age prediction described in the embodiment.

[0123] In the methods, programs, systems, devices, etc. of the embodiments of the present invention, they can be executed or implemented in a single or multiple networked computers, and can also be practiced in a distributed computing environment. In the embodiments of this specification, in these distributed computing environments, tasks can be executed by remote processing devices connected through a communication network.

[0124] Those skilled in the art should understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, those skilled in the art can think that the implementation of the functional modules / units or controllers and related method steps clarified in the above embodiments can be achieved in a manner combining software, hardware, and soft / hardware.

[0125] Unless explicitly stated, the actions or steps of the methods and programs recorded according to the embodiments of the present invention do not necessarily have to be executed in a specific order and can still achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0126] In this article, multiple embodiments of the present invention have been described. For the sake of brevity, the descriptions of each embodiment are not exhaustive, and the same or similar features or parts between the various embodiments may be omitted. In this article, "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" mean applicable to at least one embodiment or example according to the present invention, rather than all embodiments. The above terms do not necessarily mean referring to the same embodiment or example. Without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0127] The exemplary systems and methods of the present invention have been specifically shown and described with reference to the above embodiments, which are only examples of the best modes for implementing the systems and methods. Those skilled in the art can understand that various changes can be made to the embodiments of the systems and methods described here when implementing the systems and / or methods without departing from the spirit and scope of the present invention defined in the appended claims.

[0128] In addition, the methods, systems, devices, and media for zircon age prediction according to the present invention can also be implemented in the following ways:

[0129] (1) A method for zircon age prediction, characterized by including:

[0130] Collect and organize the cathodoluminescence (CL) images of zircons. The zircon characteristic information in the CL images includes at least the growth zoning characteristics of zircons.

[0131] Use a preset target zircon age prediction model to perform prediction processing on the zircon CL images and obtain the zircon age prediction results.

[0132] The target zircon age prediction model includes a zoning extraction module, a multi-scale feature fusion module, and an age prediction module. The zoning extraction module is used to extract the growth zoning characteristics from the CL image information. The multi-scale feature fusion module is used to perform multi-scale feature extraction and fusion on the growth zoning characteristics to obtain multi-scale spatial information. The age prediction module is used to obtain the final zircon age prediction results based on the multi-scale spatial information.

[0133] (2) According to the method described in (1), before the target zircon age prediction model predicts the age of zircons, it further includes training and obtaining the target zircon age model. The training and obtaining of the target zircon age model includes:

[0134] Collect and organize the CL images of the zircons to be tested, as well as the age labels and error labels of the zircons to be tested. The zircon characteristic information in the CL images includes at least the growth zoning characteristics of zircons.

[0135] Make training datasets, validation datasets, and test datasets from the CL images of the zircons to be tested, as well as the age labels and error labels.

[0136] Construct an initial zircon age prediction model.

[0137] Based on the training datasets and the validation datasets, use the stochastic gradient descent method to optimize the initial zircon age prediction model to obtain an optimized zircon age prediction model.

[0138] Use the optimized zircon age prediction model to predict the zircon CL images in the test datasets and obtain the predicted age results.

[0139] Based on the comparison and analysis of the predicted age results with the corresponding true age labels and error labels in the test datasets, obtain the target zircon age prediction model.

[0140] (3) According to the method described in (1), obtaining the optimized zircon age prediction model includes:

[0141] Use a stochastic gradient descent optimizer to train the initial zircon age prediction model; and

[0142] Use a data augmentation strategy to optimize the initial zircon age prediction model.

[0143] (4) According to the method described in (1), it is characterized in that the CL image is input into the target zircon age prediction model according to a preset size;

[0144] In the zircon zoning extraction module, input the CL image of the zircon, crop and normalize the image according to a preset size to obtain segmented image patches, input the image patches into the structure of the ViT model, process layer by layer, calculate the feature relationship through the self-attention mechanism, and perform hybrid calculation on the output features of the ViT model through convolution and multi-layer perceptron modules to obtain growth zoning features;

[0145] In the multi-scale feature fusion module, layer-by-layer extraction of the growth zoning features is performed through the transformer block, the correlation between local and global features is calculated through the self-attention mechanism, and then non-linear transformation of the feature map is performed through the feed-forward neural network. The image patches of different scales are integrated through patch merging, and the output features of different layers are stitched and fused to integrate multi-scale spatial information;

[0146] In the age prediction module, the time series information of the multi-scale spatial information is captured through the long short-term memory network, dimensionality reduction and non-linear mapping are performed using the multi-layer perceptron to generate an initial age prediction result, and the dual ranking loss is used to optimize the predicted value to rank the relative time order between samples, and the final zircon age prediction result is output.

[0147] (5) According to the method described in (4), it is characterized in that obtaining the growth zoning features includes:

[0148]

[0149] F out = MLP(Conv(F Attention , W conv ));

[0150] In the formula, where Q, K, and V represent the query vector, key vector, and value vector respectively, d k is the dimension of the key, F Attention is the output feature of the self-attention mechanism, and W conv is the convolution weight.

[0151] (6) According to the method described in (4), it is characterized in that obtaining the multi-scale spatial information includes:

[0152] F concat = [F trans1 , F trans2 , F trans3 , F trans4 ;

[0153] F unified = MLP(F concat );

[0154] F up = Upsample(F unified );

[0155] Wherein, F concat is the feature map obtained by stitching the features of F trans1 , F trans2 , F trans3 , F trans4 . F unified is the multi-scale feature fusion feature, and F up is the high-resolution feature map after upsampling.

[0156] (7) According to the method described in (4), it is characterized in that obtaining the final zircon age prediction result includes:

[0157] F time = LSTM(F up );

[0158] Age pred = MLP(F time );

[0159]

[0160] Wherein, F up is the feature map after multi-scale feature fusion, F time is the time series feature generated by LSTM, and Age pred is the age prediction value, and the prediction value is optimized using Pairwise RankingLoss.

[0161] (8) An apparatus for zircon age prediction, characterized in that according to the method for zircon age prediction described in any one of (1) to (14), the apparatus includes:

[0162] An acquisition unit, configured to acquire and organize the cathodoluminescence (CL) images of zircons, and the zircon feature information in the CL images includes at least the growth zoning features of zircons;

[0163] The zircon age prediction unit is configured to perform prediction processing on the zircon CL image by using a preset target zircon age prediction model and obtain an age prediction result of the zircon. Among them, the target zircon age prediction model includes an annulus extraction module, a multi-scale feature fusion module, and an age prediction module. The annulus extraction module is used to extract the growth annulus feature in the CL image information. The multi-scale feature fusion module is used to perform multi-scale feature extraction and fusion on the growth annulus feature to obtain multi-scale spatial information. The age prediction module is used to obtain the final zircon age prediction result based on the multi-scale spatial information.

[0164] (9) The device according to (8), wherein the zircon age prediction unit is configured to, before the target zircon age prediction model performs age prediction on the zircon, further include training and obtaining the target zircon age model. The training and obtaining the target zircon age model includes:

[0165] Collect and organize the CL images of the zircon to be tested, as well as the age labels and error labels of the zircon to be tested. The zircon feature information in the CL image at least includes the growth annulus feature of the zircon.

[0166] Make a training data set, a validation data set, and a test data set from the CL images of the zircon to be tested, the age labels, and the error labels.

[0167] Construct an initial zircon age prediction model.

[0168] Based on the training data set and the validation data set, use the stochastic gradient descent method to optimize the initial zircon age prediction model to obtain an optimized zircon age prediction model.

[0169] Use the optimized zircon age prediction model to predict the zircon CL images in the test data set to obtain a predicted age result.

[0170] Based on the comparison and analysis of the predicted age result with the corresponding true age label and error label in the test data set, obtain the target zircon age prediction model.

[0171] (10) The device according to (8), wherein the zircon age prediction unit is configured to obtain the optimized zircon age prediction model, including:

[0172] Use the stochastic gradient descent optimizer to train the initial zircon age prediction model; and

[0173] Use the data augmentation strategy to optimize the initial zircon age prediction model.

[0174] (11) The device according to (8), characterized in that the zircon age prediction unit is configured to input the CL image into a target zircon age prediction model according to a preset size;

[0175] In the zircon zoning extraction module, input the CL image of the zircon, crop and normalize the image according to a preset size to obtain segmented image patches, input the image patches into the structure of the ViT model, process layer by layer, calculate the feature relationship through the self-attention mechanism, and perform hybrid calculation on the output features of the ViT model through convolution and a multi-layer perceptron module to obtain growth zoning features;

[0176] In the multi-scale feature fusion module, layer-by-layer extraction of the growth zoning features is performed through a transformer block, the correlation between local and global features is calculated through the self-attention mechanism, and then the feature map is non-linearly transformed through a feed-forward neural network. The image patches of different scales are integrated through patch merging, and the output features of different layers are stitched and fused to integrate multi-scale spatial information;

[0177] In the age prediction module, the time series information of the multi-scale spatial information is captured through a long short-term memory network, dimensionality reduction and non-linear mapping are performed using a multi-layer perceptron to generate an initial age prediction result, and the prediction value is optimized using a pairwise ranking loss to rank the relative time order between samples, and the final zircon age prediction result is output.

[0178] (12) The device according to (8), characterized in that the zircon age prediction unit is configured to obtain the growth zoning features, including:

[0179]

[0180] F out = MLP(Conv(F Attention ,W conv ));

[0181] In the formula, where Q, K, and V respectively represent the query vector, key vector, and value vector, d k is the dimension of the key, F Attention is the output feature of the self-attention mechanism, and W conv is the convolution weight.

[0182] (13) The device according to (8), characterized in that the zircon age prediction unit is configured to obtain the multi-scale spatial information, including:

[0183] F concat = [F trans1 ,F trans2 ,F trans3,F trans4 ;

[0184] F unified = MLP(F concat );

[0185] F up = Upsample(F unified );

[0186] Wherein, F concat is the feature map obtained by feature splicing through F trans1 ,F trans2 ,F trans3 ,F trans4 , F unified is the multi-scale feature fusion feature, and F up is the high-resolution feature map after upsampling.

[0187] (14) The device according to (8), characterized in that the zircon age prediction unit is configured to obtain the final zircon age prediction result, including:

[0188] F time = LSTM(F up );

[0189] Age pred = MLP(F time );

[0190]

[0191] Wherein, F up is the feature map after multi-scale feature fusion, F time is the time series feature generated by LSTM, and Age pred is the age prediction value, and the prediction value is optimized using Pairwise RankingLoss.

[0192] (15) An electronic device, characterized in that the electronic device includes:

[0193] At least one processor; and,

[0194] A memory communicatively connected to the at least one processor; wherein,

[0195] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for zircon age prediction described in any one of (1) to (14).

[0196] (16) A non-transitory computer-readable storage medium, characterized in that the non-transitory computer-readable storage medium stores computer instructions for causing the computer to execute the method for zircon age prediction according to any one of (1) to (14).

[0197] (17) A computer program product, the computer program product comprising a computing program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions which, when executed by a computer, cause the computer to execute the method for zircon age prediction according to any one of (1) to (14).

Claims

1. A method for zircon age prediction, characterized in that, Including: Collect and organize the cathodoluminescence (CL) images of zircons, where the characteristic information of zircons in the CL images includes at least the growth zoning characteristics of zircons; Use a preset target zircon age prediction model to perform prediction processing on the CL images of zircons to obtain the age prediction results of zircons; The target zircon age prediction model includes a zoning extraction module, a multi-scale feature fusion module, and an age prediction module. The zoning extraction module is used to extract the growth zoning characteristics from the CL image information. The multi-scale feature fusion module is used to perform multi-scale feature extraction and fusion on the growth zoning characteristics to obtain multi-scale spatial information. The age prediction module is used to obtain the age prediction results of zircons based on the multi-scale spatial information.

2. The method according to claim 1, wherein Before using the target zircon age prediction model to predict the age of zircons, it also includes training and obtaining the target zircon age model. The training and obtaining of the target zircon age model includes: Collect and organize the CL images of zircons to be tested and the age labels of zircons to be tested. The characteristic information of zircons in the CL images includes at least the growth zoning characteristics of zircons; Make training datasets, validation datasets, and test datasets from the CL images of zircons to be tested and the age labels; Construct an initial zircon age prediction model; Based on the training dataset and the validation dataset, use the stochastic gradient descent method to optimize the initial zircon age prediction model to obtain an optimized zircon age prediction model; Use the optimized zircon age prediction model to predict the CL images of zircons in the test dataset to obtain the predicted age results; Based on the comparison and analysis of the predicted age results and the corresponding true age labels in the test dataset, obtain the target zircon age prediction model.

3. The method according to claim 2, characterized in that, Obtaining the optimized zircon age prediction model includes: Training the initial zircon age prediction model using a stochastic gradient descent optimizer; and Optimizing the initial zircon age prediction model using a data augmentation strategy.

4. The method according to claim 1, wherein Input the CL image into the target zircon age prediction model according to a preset size; In the zircon zoning extraction module, input the CL image of the zircon, crop and normalize the image according to a preset size to obtain segmented image patches, input the image patches into the structure of the ViT model, process layer by layer, calculate the feature relationship through the self-attention mechanism, and perform hybrid calculation on the output features of the ViT model through convolution and a multi-layer perceptron module to obtain the growth zoning characteristics; In the multi-scale feature fusion module, layer-by-layer extract the growth zoning characteristics through the transformer block, calculate the correlation between local and global features through the self-attention mechanism, perform non-linear transformation on the feature map through a feed-forward neural network, integrate image patches of different scales through patch merging, and integrate and fuse the output features of different layers to obtain multi-scale spatial information; In the age prediction module, the time series information of the multi-scale spatial information is captured by a long short-term memory network, and a multi-layer perceptron is used for dimensionality reduction and non-linear mapping to generate an initial age prediction result. The dual ranking loss is used to optimize the predicted value, sort the relative time order between samples, and output the final zircon age prediction result.

5. The method according to claim 4, wherein Obtaining the growth zoning characteristics includes: F out = MLP(Conv(F Attention , W conv )); In the formula, Q, K, and V represent the query vector, key vector, and value vector respectively, and d k is the dimension of the key, and F Attention is the output feature of the self-attention mechanism, and W conv is the convolutional weight.

6. The method according to claim 4, characterized in that Obtaining the multi-scale spatial information includes: F concat = [F trans1 , F trans2 , F trans3 , F trans4 ; F unified = MLP(F concat ); F up = Upsample(F unified ); Wherein, F concat is the feature map obtained by splicing the features of F trans1 , F trans2 , F trans3 , F trans4 ; F unified is the multi-scale feature fusion feature; F up is the high-resolution feature map after upsampling.

7. The method according to claim 4, characterized in that Obtaining the final zircon age prediction result includes: F time = LSTM(F up ); Age pred = MLP(F time ) where F up is the feature map after multi-scale feature fusion, and F time is the time series feature generated by LSTM, and Age pred is the age prediction value, and the prediction value is optimized using Pairwise RankingLoss.

8. An apparatus for zircon age prediction, characterized in that, Based on the method for zircon age prediction according to any one of claims 1-7, the device includes: An acquisition unit configured to collect and organize the cathodoluminescence (CL) images of zircons, where the zircon characteristic information in the CL images includes at least the growth zoning characteristics of zircons. A zircon age prediction unit configured to perform prediction processing on the zircon CL images using a preset target zircon age prediction model and obtain the age prediction result of zircons. Among them, the target zircon age prediction model includes an annulus extraction module, a multi-scale feature fusion module, and an age prediction module. The annulus extraction module is used to extract the growth zoning characteristics from the CL image information, the multi-scale feature fusion module is used to perform multi-scale feature extraction and fusion on the growth zoning characteristics to obtain multi-scale spatial information, and the age prediction module is used to obtain the final zircon age prediction result based on the multi-scale spatial information.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; where, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for zircon age prediction according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing the computer to execute the method for zircon age prediction according to any one of claims 1 to 7.

Images