Method and system for improving laser-induced plasma air pressure detection accuracy
By combining image processing and spectral analysis, laser-induced plasma technology extracts and processes the image and spectral information of plasma, the problem of low detection accuracy and accuracy in the prior art is solved, and high-precision online detection of the vacuum degree of vacuum switch is achieved.
Patent Information
- Application Number
- CN202411443563.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-16
- Publication Date
- 2025-03-14
AI Technical Summary
The existing laser-induced plasma gas pressure detection methods are not very accurate and accurate under complex gas scenes and environmental interference, and it is difficult to realize online detection of vacuum switch vacuum.
By combining image processing and spectral analysis, laser-induced plasma technology is used to extract two-dimensional images and one-dimensional spectral information of plasma, multi-head self-attention mechanism and ShuffleNet_V2 model are used for feature extraction, and spectral feature extraction is performed through Bi-LSTM model, and finally air pressure prediction is performed through a full connection function.
It realizes high-precision prediction of vacuum degree of vacuum switch, with detection limits up to 10-3Pa-10-4Pa, meeting the online detection requirements for vacuum degree in industrial applications and improving the accuracy and reliability of detection.
Smart Images

Figure CN119622301A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of laser induced plasma gas pressure detection, and in particular to a method and system for improving the accuracy of laser induced plasma gas pressure detection. Background Art
[0002] In the existing technology, online detection of vacuum degree of vacuum switch has always been a challenging problem. Traditional detection methods mainly rely on sensors or measuring instruments, which often have limited detection limits of only 0.1Pa-10Pa, and cannot achieve safe and reliable live detection of vacuum degree of vacuum switch, and cannot meet the real-time monitoring requirements of vacuum degree of vacuum switch in industrial applications.
[0003] The vacuum degree detection method of vacuum switch based on laser induced plasma technology is a promising new technology. This method uses the interaction between laser beam and gas to generate plasma, and infers the gas pressure by measuring the parameters of plasma. Compared with traditional methods, the vacuum degree detection method based on laser induced plasma technology has the advantages of non-contact, high sensitivity, and no need for external calibration gas, and has gradually become a research hotspot in the field of vacuum degree detection.
[0004] However, there are still some problems with the existing laser-induced plasma pressure detection methods. Traditional feature extraction and classification methods often rely on manually designed features and classifiers, and have limited adaptability to complex gas scenes and environmental interference. At the same time, traditional methods often only use a single characteristic parameter, such as the integral of the plasma's radiation intensity and the area of the radiation profile. This method cannot fully utilize other information that can be used to characterize the vacuum degree, resulting in low detection precision and accuracy.
[0005] With the development of machine learning algorithms, vacuum degree detection methods based on laser-induced plasma technology have been improved. Machine learning algorithms can improve the accuracy of vacuum degree prediction by learning a large amount of data. Compared with traditional methods, machine learning algorithms are more flexible and can automatically learn and adjust model parameters, thereby achieving more accurate air pressure prediction results.
[0006] In addition, the fusion of plasma images and spectral information is also the key to improving the accuracy of gas pressure prediction. By making full use of the information from both, the vacuum state of the vacuum switch can be more comprehensively described, and the accuracy and reliability of the detection results can be improved. Summary of the invention
[0007] In view of the above existing problems, the present invention is proposed.
[0008] Therefore, the present invention provides a method for improving the accuracy of laser-induced plasma pressure detection, which can combine image processing and spectral analysis to improve the accuracy of laser-induced plasma pressure detection, and the detection limit can reach 10-3Pa-10-4Pa, which can realize safe and reliable non-contact vacuum switch vacuum degree live detection. Laser-induced plasma technology is used to realize safe and reliable vacuum switch vacuum degree online detection, providing a high-precision and high-reliability vacuum degree monitoring solution for industrial applications.
[0009] To solve the above technical problems, the present invention provides the following technical solutions: a method for improving the accuracy of laser induced plasma pressure detection, comprising: inputting a two-dimensional image of laser plasma into an image module; performing feature extraction on the plasma in the image module and dividing it into image blocks, and embedding the image blocks into position codes; applying a multi-head self-attention mechanism to the encoded image block sequence, learning global and local information, and extracting features, and inputting the extracted features into a ShuffleNet_V2 model for feature extraction again; performing preliminary feature extraction on a one-dimensional spectrum of the laser plasma input into a spectral module, and inputting the extracted features into a Bi-LSTM model for feature extraction again; fusing the features output by the image module and the spectral module, and predicting the pressure through a fully connected function.
[0010] As a preferred solution of the method for improving the accuracy of laser induced plasma gas pressure detection described in the present invention, the image module includes extracting features through a convolutional neural network and using a self-attention mechanism to learn the global and local information of the input features.
[0011] As a preferred solution of the method for improving the accuracy of laser induced plasma gas pressure detection described in the present invention, the feature extraction again includes that the image module further extracts features extracted by the multi-head self-attention mechanism through the ShuffleNet_V2 model.
[0012] As a preferred solution of the method for improving the accuracy of laser induced plasma gas pressure detection described in the present invention, the spectral module includes feature extraction through a bidirectional long short-term memory network Bi-LSTM.
[0013] As a preferred solution of the method for improving the accuracy of laser induced plasma gas pressure detection described in the present invention, the feature fusion includes that the results of image module and spectral module classification are subjected to independent classification processes, and then fusion methods such as stacking and flattening are used to improve the accuracy and stability of gas pressure prediction.
[0014] As a preferred solution of the method for improving the accuracy of laser induced plasma gas pressure detection described in the present invention, the classification results of the image module and the spectral module include stacking and flattening the above classification results and inputting them into a fully connected function for gas pressure prediction.
[0015] As a preferred solution of the method for improving the accuracy of laser induced plasma gas pressure detection described in the present invention, wherein: the gas pressure prediction by the fully connected function includes, the classification results of the image module and the spectrum module are subjected to independent classification processes, and are input into a fully connected function through fusion to perform gas pressure prediction, the fully connected function adopts the Tanh function, the parameters are default parameters, the fully connected function matches the features extracted by the model with the labels one by one, and outputs the prediction results of the model;
[0016] When the classification confidence of the image module and the spectral module are both higher than 70%, the classification results are stacked and flattened; when the classification confidence of the image and spectral modules are both higher than 90%, the classification results are fused;
[0017] The classification result of the image module is ClassImage and the classification result of the spectral module is ClassSpectra. Stacking these two results is:
[0018] StackedOutput=Concatenate([ClassImage,ClassSpectra])
[0019] Flattened stacked features:
[0020] FlattenedOutput=Flatten(StackedOutput)
[0021] The flattened features are input into a fully connected layer and the Tanh function is used as the activation function:
[0022] DenseOutput=Dense(units=num c lasses,activation='linear')(FlattenedOutput)
[0023] TanhOutput=Tanh(DenseOutput)
[0024] Since the output range of the Tanh function is [-1,1], the output is converted into a probability distribution of air pressure prediction:
[0025] ProbabilityOutput=softmax(TanhOutput)
[0026] Among them, ProbabilityOutput is a two-dimensional tensor, which represents the model's predicted probability for each category;
[0027] When the label is binary, the probability distribution output by the model is directly used as the prediction result; when the label is continuous, the probability distribution output by the model is converted into a specific air pressure value and the prediction result of the model is output.
[0028] Another object of the present invention is to provide a system for improving the accuracy of laser induced plasma gas pressure detection, which can realize real-time monitoring of the vacuum degree of the vacuum switch and meet the demand for online detection of the vacuum degree in industrial applications.
[0029] As a preferred solution of the system for improving the accuracy of laser induced plasma gas pressure detection described in the present invention, it includes: an image module, a multi-head self-attention mechanism module, a ShuffleNet_V2 model module, a spectrum module, and a fully connected function module;
[0030] The image module extracts features through a convolutional neural network, divides the image into image blocks, and embeds a position code for each image block;
[0031] The multi-head self-attention mechanism module applies the multi-head self-attention mechanism to learn the global and local information in the image block sequence and extract features;
[0032] The ShuffleNet_V2 model module uses the ShuffleNet_V2 model to perform feature extraction again on the extracted features;
[0033] The spectrum module performs preliminary feature extraction on the spectrum and inputs the extracted features into the Bi-LSTM model for further feature extraction;
[0034] The fully connected function module performs independent classification processes on the classification results of the image module and the spectrum module, and then fuses them into a fully connected function to perform air pressure prediction.
[0035] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, a method for improving the accuracy of laser induced plasma gas pressure detection is implemented.
[0036] A computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, steps of a method for improving the accuracy of laser-induced plasma gas pressure detection are implemented.
[0037] Beneficial effects of the invention: This method combines advanced technologies such as laser induced plasma technology, image processing and spectral analysis, and can make full use of the two-dimensional image and one-dimensional spectral information of the plasma to achieve high-precision prediction of the vacuum degree of the vacuum switch. Compared with traditional methods, the method disclosed in this invention has higher prediction accuracy and stability, and the detection limit can reach 10 -3 Pa–10 -4 Pa. Traditional air pressure detection methods often rely on sensors or measuring instruments, making it difficult to achieve online detection of the vacuum degree of vacuum switches. This method is based on laser induced plasma technology and uses a non-contact detection method to achieve real-time monitoring of the vacuum degree of vacuum switches, meeting the needs for online detection of vacuum degree in industrial applications. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:
[0039] Figure 1 A schematic flow chart of a method for improving the accuracy of laser induced plasma gas pressure detection provided by an embodiment of the present invention.
[0040] Figure 2 A schematic diagram of a model framework of a method for improving the accuracy of laser induced plasma gas pressure detection provided by an embodiment of the present invention.
[0041] Figure 3 A confusion matrix of prediction results of a method for improving the accuracy of laser induced plasma gas pressure detection provided by an embodiment of the present invention.
[0042] Figure 4 A schematic diagram of working modules of a system for improving the accuracy of laser induced plasma gas pressure detection provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0043] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.
[0044] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0045] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.
[0046] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.
[0047] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.
[0048] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0049] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides a method for improving the accuracy of laser induced plasma pressure detection, comprising:
[0050] S1: Input the two-dimensional image of the laser plasma into the image module, extract the features of the plasma in the image module and segment it into image blocks, and embed the image blocks into position codes.
[0051] Furthermore, the image module performs feature extraction through a convolutional neural network, which utilizes the Vision Transformer architecture to divide the plasma image into several image blocks. Each image is position-encoded separately, and the global and local information of the input features are learned through a multi-head self-attention mechanism. The self-attention mechanism allows each input to interact with each other, and then finds the inputs that they should pay more attention to. The output of the self-attention module is the aggregation and attention score of these interactions.
[0052] S2: Apply the multi-head self-attention mechanism to the encoded image block sequence, learn global and local information, and extract features. The extracted features are input into the ShuffleNet_V2 model for feature extraction again.
[0053] The image module further extracts features from the multi-head self-attention mechanism through the ShuffleNet_V2 model. The ShuffleNet_V2 model is an efficient image classification model. The features extracted from the multi-head self-attention mechanism are input into the ShuffleNet_V2 model. The model parameters are the default parameters. After the model runs, the extracted features are output.
[0054] S3: Perform preliminary feature extraction on the one-dimensional spectrum of the laser plasma input into the spectrum module, and input the extracted features into the Bi-LSTM model for feature extraction again.
[0055] Furthermore, the spectral module performs feature extraction through a bidirectional long short-term memory network (Bi-LSTM), which is composed of a forward long short-term memory network and a backward long short-term memory network; the parameters of the Bi-LSTM are default parameters, the spectral features are input into the Bi-LSTM, and then the model will output the spectral features.
[0056] S4: The output features of the image module and the spectral module are fused and the air pressure is predicted through the fully connected function.
[0057] The results of both image module and spectral module classification go through independent classification processes and then go through fusion methods, such as stacking and flattening, to improve the accuracy and stability of air pressure prediction.
[0058] The classification results of the image module and the spectral module are stacked and flattened, and then input into a fully connected function for air pressure prediction. The fully connected function uses the Tanh function with default parameters. The fully connected function matches the features extracted by the model with the labels one by one and outputs the prediction results of the model.
[0059] Furthermore, the air pressure prediction by the fully connected function includes: the classification results of the image module and the spectrum module are subjected to independent classification processes, and then fused and input into a fully connected function for air pressure prediction. The fully connected function adopts the Tanh function, and the parameters are default parameters. The fully connected function matches the features extracted by the model with the labels one by one, and outputs the prediction results of the model;
[0060] When the classification confidence of the image module and the spectral module are both higher than 70%, the classification results are stacked and flattened; when the classification confidence of the image and spectral modules are both higher than 90%, the classification results are fused;
[0061] The classification result of the image module is ClassImage and the classification result of the spectral module is ClassSpectra. Stacking these two results is:
[0062] StackedOutput=Concatenate([ClassImage,ClassSpectra])
[0063] Flattened stacked features:
[0064] FlattenedOutput=Flatten(StackedOutput)
[0065] The flattened features are input into a fully connected layer and the Tanh function is used as the activation function:
[0066] DenseOutput=Dense(units=num c lasses,activation='linear')(FlattenedOutput)
[0067] TanhOutput=Tanh(DenseOutput)
[0068] Since the output range of the Tanh function is [-1,1], the output is converted into a probability distribution of air pressure prediction:
[0069] ProbabilityOutput=softmax(TanhOutput)
[0070] Among them, ProbabilityOutput is a two-dimensional tensor, which represents the model's predicted probability for each category;
[0071] When the label is binary, the probability distribution output by the model is directly used as the prediction result; when the label is continuous, the probability distribution output by the model is converted into a specific air pressure value and the prediction result of the model is output.
[0072] Example 2, reference Figure 3 , which is an embodiment of the present invention, provides a method for improving the accuracy of laser induced plasma gas pressure detection. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.
[0073] The sample set contains 10 5 Pa, 10 4 Pa, 10 3 Pa, 10 2 Pa, 10Pa, 1Pa, 10 -1 Pa, 10 -2 Pa, 10 -3 Pa, 10 -4 Two-dimensional plasma image samples and one-dimensional plasma spectrum samples collected at ten gas pressures such as Pa, with 120 image samples and 120 spectrum samples at each gas pressure, a total of 1200 two-dimensional plasma image samples and 1200 one-dimensional plasma spectrum samples;
[0074] Preferably, in another embodiment of the present disclosure, a five-fold cross validation training is performed on the above sample set, and the air pressure prediction result of the method of the present disclosure is as follows: Figure 3 As shown in Table 1, the method for improving the accuracy of laser induced plasma pressure detection in the present disclosure is 10 5 Pa-10 -2 Pa pressure range, the pressure prediction accuracy is 100%, the pressure is 10 -3 The accuracy is 95.00% when the air pressure is 10 -4 Pa, the accuracy rate is 73.33%; it can be seen that the method for improving the accuracy of laser induced plasma gas pressure detection disclosed in the present invention can efficiently and accurately predict the gas pressure value.
[0075] Table 1
[0076]
[0077] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
[0078] Embodiment 3, the third embodiment of the present invention, is different from the first two embodiments in that:
[0079] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.
[0080] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.
[0081] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.
[0082] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0083] Example 4, reference Figure 4 , which is an embodiment of the present invention, provides a system for improving the accuracy of laser induced plasma gas pressure detection, characterized by: including an image module 1, a multi-head self-attention mechanism module 2, a ShuffleNet_V2 model module 3, a spectrum module 4, and a fully connected function module 5.
[0084] Image module 1 performs feature extraction through a convolutional neural network, divides the image into image blocks, and embeds a position code for each image block.
[0085] Multi-head self-attention mechanism module 2 applies the multi-head self-attention mechanism to learn the global and local information in the image block sequence and extract features.
[0086] ShuffleNet_V2 model module 3, uses the ShuffleNet_V2 model to perform feature extraction on the extracted features again.
[0087] Spectral module 4 performs preliminary feature extraction on the spectrum and inputs the extracted features into the Bi-LSTM model for further feature extraction.
[0088] The fully connected function module 5 inputs the classification results of the image module and the spectrum module into a fully connected function for air pressure prediction after an independent classification process.
[0089] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A method for improving the accuracy of laser-induced plasma pressure detection, characterized in that: include, Inputting a two-dimensional image of the laser plasma into an image module; In the image module, the plasma is feature extracted and segmented into image blocks, and the image blocks are embedded in the position code; Apply a multi-head self-attention mechanism to the encoded image block sequence to learn global and local information and extract features. The extracted features are input into the ShuffleNet_V2 model for feature extraction again. Perform preliminary feature extraction on the one-dimensional spectrum of the laser plasma input into the spectrum module, and input the extracted features into the Bi-LSTM model for feature extraction again; The output features of the image module and the spectral module are fused and the air pressure is predicted through the fully connected function.
2. A method for improving the accuracy of laser induced plasma gas pressure detection according to claim 1, characterized in that: The image module includes extracting features through a convolutional neural network and learning the global and local information of input features using a self-attention mechanism.
3. A method for improving the accuracy of laser induced plasma gas pressure detection as claimed in claim 2, characterized in that: The re-extraction of features includes that the image module re-extracts features extracted by the multi-head self-attention mechanism through the ShuffleNet_V2 model.
4. A method for improving the accuracy of laser induced plasma gas pressure detection as claimed in claim 3, characterized in that: The spectral module includes feature extraction through a bidirectional long short-term memory network Bi-LSTM.
5. A method for improving the accuracy of laser induced plasma gas pressure detection as claimed in claim 4, characterized in that: The feature fusion includes that the classification results of the image module and the spectral module are subjected to independent classification processes and then fused by a fusion method, such as stacking and flattening, to improve the accuracy and stability of air pressure prediction.
6. A method for improving the accuracy of laser induced plasma gas pressure detection according to claim 5, characterized in that: The classification results of the image module and the spectrum module include stacking and flattening the above classification results and inputting them into a fully connected function for air pressure prediction.
7. A method for improving the accuracy of laser induced plasma pressure detection according to claim 6, characterized in that: The air pressure prediction by the fully connected function includes: the classification results of the image module and the spectrum module are subjected to independent classification processes, and then fused and input into a fully connected function for air pressure prediction. The fully connected function adopts the Tanh function, and the parameters are default parameters. The fully connected function matches the features extracted by the model with the labels one by one, and outputs the prediction results of the model; When the classification confidence of the image module and the spectral module are both higher than 70%, the classification results are stacked and flattened; when the classification confidence of the image and spectral modules are both higher than 90%, the classification results are fused; The classification result of the image module is ClassImage and the classification result of the spectral module is ClassSpectra. Stacking these two results is: StackedOutput=Concatenate([ClassImage,ClassSpectra]) Flattened stacked features: FlattenedOutput=Flatten(StackedOutput) The flattened features are input into a fully connected layer and the Tanh function is used as the activation function: DenseOutput=Dense(units=num c lasses,activation=′linear′)(FlattenedOutput) TanhOutput=Tanh(DenseOutput) Since the output range of the Tanh function is [-1,1], the output is converted into a probability distribution of air pressure prediction: ProbabilityOutput=softmax(TanhOutput) Among them, ProbabilityOutput is a two-dimensional tensor, which represents the model's predicted probability for each category; When the label is binary, the probability distribution output by the model is directly used as the prediction result; when the label is continuous, the probability distribution output by the model is converted into a specific air pressure value and the prediction result of the model is output.
8. A system using the method for improving the accuracy of laser induced plasma pressure detection as claimed in any one of claims 1 to 7, characterized in that: Including image module, multi-head self-attention mechanism module, ShuffleNet_V2 model module, spectrum module, and fully connected function module; The image module extracts features through a convolutional neural network, divides the image into image blocks, and embeds a position code for each image block; The multi-head self-attention mechanism module applies the multi-head self-attention mechanism to learn the global and local information in the image block sequence and extract features; The ShuffleNet_V2 model module uses the ShuffleNet_V2 model to perform feature extraction again on the extracted features; The spectrum module performs preliminary feature extraction on the spectrum and inputs the extracted features into the Bi-LSTM model for further feature extraction; The fully connected function module performs independent classification processes on the classification results of the image module and the spectrum module, and then fuses them into a fully connected function to perform air pressure prediction.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.
