Rapid bacillus plant type identification method based on Raman spectrum and deep learning

Abstract

The invention discloses a bacillus plant type rapid identification method based on Raman spectrum and deep learning, and relates to the technical field of microbiological detection.The method comprises the steps that Raman spectrum data of bacillus to be detected are obtained, and the Raman spectrum data are collected in a Raman spectrum fingerprint section and subjected to standardized preprocessing; then, the Raman spectrum data are input into a trained RamanFormer model, a plant type prediction result is obtained, and the RamanFormer model is obtained through training of a training set containing the Raman spectrum data of various known plant type bacillus and corresponding plant type labels; and finally, outputting a plant type identification result of the bacillus to be detected. According to the method, the Raman spectrum fingerprint region data of the bacillus is collected and subjected to standardized treatment, the plant type specific characteristics are extracted by utilizing the RamanFormer model, and classification is completed, so that the technical problem that the traditional bacillus plant type identification method is low in related plant type distinguishing accuracy is solved, and rapid and accurate identification of the bacillus related plant types is realized.

Description

Technical Field

[0001] This application relates to the field of microbial detection technology, and in particular to a rapid identification method for spore-forming baculotype strains based on Raman spectroscopy and deep learning. Background Technology

[0002] Raman spectroscopy, as a rapid, non-destructive, label-free, and highly specific analytical technique, has demonstrated significant technological advantages in the field of microbial detection. Raman spectroscopy enables rapid identification and typing of pathogens, for example, in screening for sepsis pathogens, detecting foodborne pathogens, and analyzing aquatic microbial communities.

[0003] However, in the field of soil microbial detection, especially in the precise typing and identification of Bacillus subtypes, current methods mainly rely on traditional methods. These traditional methods include morphological identification, molecular biological identification, and physiological and biochemical identification. Morphological identification depends on the operator's experience and judgment, is highly subjective, and has limited accuracy; molecular biological identification is complex, time-consuming, and requires specialized reagents and equipment; and physiological and biochemical identification involves cumbersome steps and long reaction cycles, making it difficult to meet the needs of rapid identification.

[0004] Although Raman spectroscopy has the potential for rapid analysis, existing spectroscopic analysis methods lack the ability to rapidly analyze soil microbial strains, especially for the rapid identification of different spore-forming rod-shaped strains that are closely related and derived from the same strain. Summary of the Invention

[0005] The purpose of this application is to provide a rapid identification method for Bacillus strains based on Raman spectroscopy and deep learning. By collecting and standardizing the Raman spectral fingerprint data of Bacillus, the RamanFormer model is used to extract strain-specific features and complete the classification. This solves the technical problem of low accuracy in distinguishing closely related strains by traditional Bacillus strain identification methods, and achieves rapid and accurate identification of closely related Bacillus strains.

[0006] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a rapid identification method for Bacillus strains based on Raman spectroscopy and deep learning, comprising: acquiring Raman spectral data of the Bacillus strain to be tested, wherein the Raman spectral data is data collected focusing on the Raman spectral fingerprint region and preprocessed by standardization; inputting the Raman spectral data into a trained RamanFormer model to obtain a strain type prediction result; wherein the RamanFormer model is constructed based on the Transformer architecture and trained on a training set containing Raman spectral data of various known strains of Bacillus and corresponding strain type labels, and is used to extract strain type specific features from the Raman spectral data; and outputting the strain type identification result of the Bacillus strain to be tested, wherein the identification result is determined based on the strain type prediction result.

[0007] Optionally, before acquiring the Raman spectral data of the Bacillus spp. to be tested, a model training step is further included; wherein, the model training step includes: acquiring a training dataset, the training dataset containing Raman spectral data of various known strains of Bacillus spp. and corresponding strain labels, wherein the Raman spectral data in the training dataset is collected focusing on the Raman spectral fingerprint segment and is preprocessed by standardization; and training an initial RamanFormer model based on the Transformer architecture using the training dataset to obtain the trained RamanFormer model.

[0008] Optionally, the multiple known strains are multiple variant strains obtained from the same Bacillus parent strain through experimental mutation.

[0009] Optionally, the number of the plurality of variant strains is four.

[0010] Optionally, in training the initial RamanFormer model based on the Transformer architecture using the training dataset: the cross-entropy function is used as the objective function, the Adam optimizer is employed, the training batch size is 16, the number of iterations is 100, and the learning rate is 1.0 × 10⁻⁶. -5 .

[0011] Optionally, the standardization preprocessing is Z-score standardization, which includes: for the Raman spectral data matrix containing all plant types, calculating the mean and standard deviation of the intensity values ​​of all samples in each wavelength dimension; and performing a normalization transformation on the intensity value of each sample at each wavelength based on the mean and the standard deviation.

[0012] Optionally, the RamanFormer model includes a feature fusion layer, at least one window attention computation layer, and a fully connected classification layer.

[0013] Optionally, the number of window attention calculation layers is three, the depths of the three window attention calculation layers are 2, 2 and 6 respectively, and the window size is 7 for each.

[0014] Secondly, this application provides a rapid identification device for Bacillus strains based on Raman spectroscopy and deep learning, comprising: a data acquisition module configured to acquire Raman spectral data of the Bacillus strain to be tested, wherein the Raman spectral data is data collected focusing on the Raman spectral fingerprint region and preprocessed by standardization; a strain identification module configured to input the Raman spectral data into a trained RamanFormer model to obtain a strain type prediction result; wherein the RamanFormer model is constructed based on the Transformer architecture and trained using a training set containing Raman spectral data of various known Bacillus strains and corresponding strain type labels, and is used to extract strain type specific features from the Raman spectral data; and an output module configured to output the strain type identification result of the Bacillus strain to be tested, wherein the identification result is determined based on the strain type prediction result.

[0015] Optionally, it also includes a model training module; the model training module is configured to acquire a training dataset, the training dataset containing Raman spectral data of various known strains of Bacillus and corresponding strain labels, the Raman spectral data in the training dataset being collected by focusing on the Raman spectral fingerprint segment and being preprocessed by standardization; the initial RamanFormer model built based on the Transformer architecture is trained using the training dataset to obtain the trained RamanFormer model.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a rapid identification method for Bacillus strains based on Raman spectroscopy and deep learning. It acquires standardized preprocessed fingerprint region Raman spectral data, which is then input into a RamanFormer model trained on a specific training set. Since the RamanFormer model is built on a Transformer architecture and trained on a training set containing Raman spectral data of various known Bacillus strains and corresponding strain labels, it extracts strain-specific features from the Raman spectral data. This overcomes the limitations of traditional convolutional neural network (CNN) models in capturing long-range dependencies in Raman spectra and mining weak feature correlations. Furthermore, traditional CNN models struggle to fully adapt to the complex feature distribution patterns of the Raman spectral fingerprint region, enabling more accurate extraction of deep spectral features. Finally, the identification result is determined and output based on the strain prediction results output by the model. This overcomes the drawbacks of traditional identification methods that rely on manual interpretation, are highly subjective, and time-consuming, achieving automated and rapid identification, and obtaining identification conclusions in a short time. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an application environment diagram of a rapid identification method for Bacillus strains based on Raman spectroscopy and deep learning in one embodiment of this application; Figure 2 This is a flowchart illustrating the method for constructing and training the RamanFormer model provided in Embodiment 1 of this application; Figure 3 This is the Raman spectrum of strain H in Example 1 of this application; Figure 4 This is the Raman spectrum of the P-type strain in Example 1 of this application; Figure 5 This is the Raman spectrum of the Y-type strain in Example 1 of this application; Figure 6 This is the Raman spectrum of the Z-type strain in Example 1 of this application; Figure 7 This is a schematic diagram of the RamanFormer model architecture in Embodiment 1 of this application; Figure 8 This is a schematic diagram of the Raman spectral data standardization preprocessing flowchart in Embodiment 1 of this application; Figure 9This is a flowchart illustrating the rapid identification method for the strain of the Bacillus spp. to be tested provided in Embodiment 2 of this application; Figure 10 A schematic diagram of the functional modules of a rapid identification device for Bacillus strains based on Raman spectroscopy and deep learning is provided in Embodiment 3 of this application; Figure 11 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] The rapid identification method for spore-forming baculotype strains based on Raman spectroscopy and deep learning provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be set up independently, integrated into server 102, or placed in the cloud or on another server. Terminal 101 can send the Raman spectral data of the Bacillus spores to be tested to server 102. Server 102 receives the Raman spectral data and inputs it into a trained RamanFormer model to obtain a strain type prediction result. The RamanFormer model is built based on the Transformer architecture and trained using a training set containing Raman spectral data of various known Bacillus strains and corresponding strain type labels. It is used to extract strain type-specific features from the Raman spectral data. Then, it outputs the strain type identification result of the Bacillus spores to be tested, which is determined based on the strain type prediction result. Server 102 can feed back the obtained strain type identification result to terminal 101. In addition, in some embodiments, the strain identification results can also be achieved by the server 102 or the terminal 101 alone. For example, the terminal 101 can directly obtain the Raman spectral data of the Bacillus spores to be tested, or the server 102 can obtain the Raman spectral data of the Bacillus spores to be tested from the data storage system.

[0022] The terminal 101 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 102 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0023] Example 1 This embodiment provides a method for constructing and training the RamanFormer model.

[0024] This embodiment uses the identification of multiple closely related strains obtained through experimental mutation of the same Bacillus parent strain as an example. In one specific example, the variant strain includes four types, which are named H-type, Y-type, P-type, and Z-type in this embodiment for ease of description. Those skilled in the art will understand that the core of this method lies in the RamanFormer model's ability to extract subtle spectral features, and it is also applicable to identifying other Bacillus strains with similar phylogenetic relationships and spectral characteristics.

[0025] like Figure 2 As shown, the method for constructing and training the RamanFormer model in this embodiment specifically includes the following steps: S11: Obtain the training dataset.

[0026] In practice, Raman spectral data of four known strains of Bacillus (H, P, Y, and Z) and their corresponding strain labels were acquired as a training dataset. These four known strains are multiple variant strains derived from the same parent strain of Bacillus, obtained through experimental mutation.

[0027] In practice, the Raman spectral data of the training dataset were obtained as follows: For the four strains (H, P, Y, and Z), Raman spectral data for each strain were measured using a Raman spectrometer after cultivation. To ensure data comparability, uniform acquisition conditions were set: laser wavelength 784.56 nm, acquisition time 25 s. Spectral data were focused on the Raman spectral fingerprint region. Specifically, the Raman spectrum of strain H is shown below. Figure 3 As shown, the Raman spectrum of the P-type strain is as follows: Figure 4 As shown, the Raman spectrum of strain Y is as follows: Figure 5 As shown, the Raman spectrum of the Z-type strain is as follows: Figure 6 As shown, in the Raman spectra of each strain, the Raman Shift on the horizontal axis represents the Raman shift, and the Intensity on the vertical axis represents the Raman intensity.

[0028] In practice, the raw spectral data collected needs to undergo standardized preprocessing. Specifically, the standardized preprocessing is Z-score standardization.

[0029] After the above collection and preprocessing, the training dataset is obtained, which stores the Raman spectral data of each sample and the corresponding plant type labels (H, P, Y, Z).

[0030] S12: Train the initial RamanFormer model built on the Transformer architecture using the training dataset.

[0031] In practice, the training dataset obtained in step S101 is input into the constructed initial RamanFormer model for training. The training process uses the cross-entropy function as the objective function, employs the Adam optimizer, and sets the training batch size to 16, the number of iterations to 100, and the learning rate to 1.0 × 10⁻⁶. -5 .

[0032] The initial RamanFormer model is built on the Transformer architecture, comprising a feature fusion layer, at least one window attention computation layer, and a fully connected classification layer. In a specific example, the number of window attention computation layers is three, with depths of 2, 2, and 6 respectively, and the window size is 7 for each. A schematic diagram of the RamanFormer model architecture can be found here. Figure 7 .

[0033] Once training is complete, a trained RamanFormer model that can be used for plant type identification is obtained.

[0034] Implementing steps S11 and S12 above, step S11 involves acquiring spectra focused on the fingerprint region under uniform acquisition conditions, and preprocessing all plant type data uniformly using Z-score normalization. This solves the problem of high noise and poor comparability in training data caused by fluctuations in spectral acquisition conditions or inconsistent data scales, providing a reliable data foundation for training a high-performance discrimination model. Step S12 employs a specific initial RamanFormer model built on the Transformer architecture, and trains it using the training dataset with specific parameters such as the cross-entropy function and the Adam optimizer, thus addressing the limitations of traditional convolutional neural networks (…). CNN models have limitations in capturing long-range dependencies in Raman spectra and mining weak feature correlations. Furthermore, traditional convolutional neural network (CNN) models struggle to fully adapt to the complex feature distribution patterns in Raman spectral fingerprint regions, thus achieving more accurate extraction of deep spectral features. In addition, the RamanFormer model architecture, which includes a feature fusion layer and three window attention calculation layers, constructed in step S12, solves the technical bottleneck of existing models' inability to simultaneously and effectively model the relationship between fine local spectral features and global context. This achieves a more accurate adaptation to the complex feature distribution patterns in Raman spectral fingerprint regions, thereby significantly improving the completeness and discriminative power of feature representation.

[0035] In specific implementation, such as Figure 8 As shown, the Z-score normalization method was used for the standardization preprocessing of Raman spectral data acquired by the Raman spectrometer. The specific steps include: S21, for the Raman spectral data matrix containing all plant type samples, calculate the mean and standard deviation of the intensity values ​​of all samples in each wavelength dimension.

[0036] S22, based on the mean and standard deviation, normalizes the intensity value of each sample at each wavelength.

[0037] The transformed data matrix was validated by calculating the mean and variance of all sample intensity values ​​for each wavelength dimension. The validation showed that the mean for each wavelength dimension approached 0, and the variance approached 1, indicating that the Z-score standardization process achieved the expected results and the data met the input requirements for subsequent models.

[0038] By implementing steps S21 and S22 above, and uniformly calculating the mean and standard deviation of the merged plant type samples, the problem of artificially weakening the inherent spectral differences between different plant types when grouping and standardizing is solved. This achieves the preservation and highlighting of key discriminative information between plant types under a unified standard. By normalizing the intensity values ​​of each wavelength of each sample based on the overall mean and standard deviation, the spectral data of all samples are transformed to a standard normal distribution scale centered at zero with unit variance. This ensures the comparability of the data and the stability of model training, and provides standardized input data for the subsequent deep learning model to extract effective plant type-specific features.

[0039] Example 2 This embodiment describes a method for rapid identification of strain types of the target Bacillus strain using the trained RamanFormer model obtained from Example 1. Figure 9 As shown, the method specifically includes: S31, Obtain Raman spectral data of the Bacillus spp. to be tested.

[0040] In the specific implementation, after culturing the *Bacillus* to be tested, Raman spectra were measured using a Raman spectrometer. To ensure data comparability, the same acquisition conditions as in Example 1, S101 were used: laser wavelength 784.56 nm, acquisition time 25 s. The Raman spectra were collected by focusing on the Raman spectral fingerprint region. The acquired data then underwent standardized preprocessing, which was the same as the Z-score normalization method described in Example 1, S101. Specifically, the mean and standard deviation calculated from the training dataset were used to normalize the intensity values ​​of each sample at each wavelength. This step yielded the standardized preprocessed data.

[0041] S32. Input the Raman spectral data into the trained RamanFormer model to obtain the plant type prediction results.

[0042] In a specific implementation, the standardized preprocessed Raman spectral data obtained in step S201 is input into the trained RamanFormer model.

[0043] S33 outputs the strain type identification results of the tested Bacillus strain. The identification results are determined based on the strain type prediction results.

[0044] In a specific implementation, based on the strain type prediction results obtained in step S32, the strain type identification result of the Bacillus strain to be tested is determined and output. In one specific implementation, multiple Raman spectral data can be measured for the same Bacillus strain to be tested, for example, about 1000 spectra. After obtaining the corresponding strain type prediction results for each, the final identification result of the strain is determined based on the multiple strain type prediction results by statistically analyzing the frequency of occurrence of each strain type.

[0045] By implementing steps S31 to S33, the standardized preprocessed fingerprint region Raman spectral data is obtained in step S31. Then, in step S32, a RamanFormer model based on the Transformer architecture and trained on a specific training set is used. This solves the technical problem of low identification accuracy in the identification of closely related strains of Bacillus species, where traditional methods struggle to capture and quantify specific features due to subtle spectral differences. The model accurately captures local correlation features and long-range dependencies in different wavelength ranges of spectral data and precisely extracts and identifies deep-seated specific features in the Raman spectrum. In step S33, the strain type prediction results based on the model output are used to determine and output the final identification result. This solves the drawbacks of traditional identification methods that rely on manual interpretation, are highly subjective, and time-consuming. The identification process is automated and rapid, and identification conclusions are obtained in a short time.

[0046] Example 3 A rapid identification device for spore-forming baicale strains based on Raman spectroscopy and deep learning is provided, such as... Figure 10 As shown, it includes: The data acquisition module 201 is configured to acquire Raman spectral data of the Bacillus spp. to be tested, wherein the Raman spectral data is data collected by focusing on the Raman spectral fingerprint region and preprocessed by standardization. The strain type identification module 202 is configured to input the Raman spectral data into a trained RamanFormer model to obtain strain type prediction results; wherein, the RamanFormer model is built based on the Transformer architecture and is trained on a training set containing Raman spectral data of various known strains of Bacillus and corresponding strain type labels, and is used to extract strain type specific features from the Raman spectral data. Output module 203 is configured to output the strain type identification result of the tested Bacillus subtilis, the identification result being determined based on the strain type prediction result; The model training module 204 is configured to acquire a training dataset containing Raman spectral data of various known strains of Bacillus and their corresponding strain labels. The Raman spectral data in the training dataset is collected by focusing on the Raman spectral fingerprint segment and is preprocessed by standardization. The initial RamanFormer model built on the Transformer architecture is trained using the training dataset to obtain the trained RamanFormer model.

[0047] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 11 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When the computer program is executed by the processor, it implements a rapid identification method for Bacillus spore strains based on Raman spectroscopy and deep learning.

[0048] Those skilled in the art will understand that Figure 11 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0049] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0050] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0051] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0052] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0053] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Furthermore, any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory.

[0054] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0055] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A rapid identification method for spore-forming baicale strains based on Raman spectroscopy and deep learning, characterized in that, include: Raman spectral data of the target Bacillus strain were obtained, wherein the Raman spectral data were collected by focusing on the Raman spectral fingerprint region and preprocessed by standardization. The Raman spectral data is input into the trained RamanFormer model to obtain the strain type prediction result; wherein, the RamanFormer model is built based on the Transformer architecture and is trained on a training set containing Raman spectral data of various known strains of Bacillus and corresponding strain type labels, and is used to extract strain type specific features from the Raman spectral data. Output the strain type identification result of the tested Bacillus species, which is determined based on the strain type prediction result.

2. The rapid identification method for spore-forming baculotype strains based on Raman spectroscopy and deep learning according to claim 1, characterized in that, Before acquiring the Raman spectral data of the target Bacillus, a model training step is included; wherein the model training step includes: A training dataset is obtained, which contains Raman spectral data of various known strains of Bacillus and their corresponding strain labels. The Raman spectral data in the training dataset is collected by focusing on the Raman spectral fingerprint segment and is preprocessed by standardization. The initial RamanFormer model, built on the Transformer architecture, is trained using the training dataset to obtain the trained RamanFormer model.

3. The rapid identification method for spore-forming baculotype strains based on Raman spectroscopy and deep learning according to claim 2, characterized in that, The various known strains are derived from the same Bacillus parent strain and are multiple variant strains obtained through experimental mutation.

4. The rapid identification method for spore-forming baculotype strains based on Raman spectroscopy and deep learning according to claim 3, characterized in that, The number of the multiple variant strains is four.

5. The rapid identification method for spore-forming baculotype strains based on Raman spectroscopy and deep learning according to claim 2, characterized in that, In the training of the initial RamanFormer model based on the Transformer architecture using the aforementioned training dataset: the cross-entropy function was used as the objective function, the Adam optimizer was employed, the training batch size was 16, the number of iterations was 100, and the learning rate was 1.0 × 10⁻⁶. -5 .

6. The rapid identification method for spore-forming baculotype strains based on Raman spectroscopy and deep learning according to claim 2, characterized in that, The standardization preprocessing is Z-score standardization, including: For the Raman spectral data matrix containing all plant types, the mean and standard deviation of the intensity values ​​of all samples under each wavelength dimension are calculated. Based on the mean and the standard deviation, the intensity value of each sample at each wavelength is normalized.

7. The rapid identification method for spore-forming baculotype strains based on Raman spectroscopy and deep learning according to claim 1, characterized in that, The RamanFormer model includes a feature fusion layer, at least one window attention calculation layer, and a fully connected classification layer.

8. The rapid identification method for spore-forming baculotype strains based on Raman spectroscopy and deep learning according to claim 7, characterized in that, The number of window attention calculation layers is 3, the depths of the 3 window attention calculation layers are 2, 2 and 6 respectively, and the window size is 7 for each layer.

9. A rapid identification device for spore-forming baicale strains based on Raman spectroscopy and deep learning, characterized in that, include: The data acquisition module is configured to acquire Raman spectral data of the Bacillus spp. to be tested, wherein the Raman spectral data is data collected by focusing on the Raman spectral fingerprint region and preprocessed by standardization. The strain type identification module is configured to input the Raman spectral data into a trained RamanFormer model to obtain strain type prediction results; wherein, the RamanFormer model is built based on the Transformer architecture and is trained on a training set containing Raman spectral data of various known strains of Bacillus and corresponding strain type labels, and is used to extract strain type specific features from the Raman spectral data. The output module is configured to output the strain type identification result of the tested Bacillus spp., the identification result being determined based on the strain type prediction result.

10. The rapid identification device for spore-forming baculotype strains based on Raman spectroscopy and deep learning according to claim 9, characterized in that, It also includes a model training module; The model training module is configured to acquire a training dataset containing Raman spectral data of various known strains of Bacillus and their corresponding strain labels. The Raman spectral data in the training dataset is collected by focusing on the Raman spectral fingerprint region and is preprocessed by standardization. The initial RamanFormer model built on the Transformer architecture is trained using the training dataset to obtain the trained RamanFormer model.

Images